Skip to main content
Internet Archive's 25th Anniversary Logo

Full text of "TRANSACTIONS OF THE AMERICAN INSTITUTE OF MINING AND METALLURGICAL ENGINEERS VOLUME 143. INSTITUTE OF METALS DIVISION 1941"

See other formats




6PQ.6 



V.143 



ifUllbtUi (filttj 




This Volume is for 
REFERENCE USE ONLY 



3Esfla^^ 



OF THE 



AMEBICAS INSTITUTE OF MIMNG 
METALLMGICAI EMOEEES 



Volume 143 



INSTITUTE OF METALS DIVISION 

1941 



PAPERS AND DISCUSSIONS PRESENTED BEFORE THE DIVISION AT MEETINGS HELD AT 
CLEVELAND, OCTOBER 21-23, 1940, AND NEW YORK FEBRUARY 17-20, 1941 



PUBLISHED BY THE INSTITUTE 

AT THE OFFICE OF THE SECRETARY 

29 WEST 3QTH STREET 

NEW YORK, N. Y. 



This volume is the fifteenth of a series constituting 
the official proceedings of the Institute of Metals Division 
of the American Institute of Mining and Metallurgical 
Engineers. It deals with nonferrous metals and includes 
papers presented at the Cleveland Meeting, Oct. 2123, 
1940, and the New York Meeting, Feb. 1720, 1941. The 
complete list of publications and proceedings, including 
the present volume, is as follows: 

1908-1911 Transactions of the American Brass Founders 7 
Association: 1908, Vols. i and 2; 1909, Vol. 3; 
1910, Vol. 4; 1911, Vol. 5. 

19121916 Transactions of the American Institute of Metals, 
Vols. 6-10. 

19171918 Journal of the American Institute of Metals, 
Vols. ii 12. 

1919-1926 TRANSACTIONS of the American Institute of 
Mining and Metallurgical Engineers, Volumes 
60, 64, 67, 68, 69, 70, 71 and 73. 

19271928 PROCEEDINGS of the Institute of Metals Divi 
sion of the American Institute of Mining and. 
Metallurgical Engineers, two volumes, of 
which the later is now designated Vol. 78 of the 
A. I. M. E. TRANSACTIONS. 

19291941 TRANSACTIONS of the American Institute of 
Mining and Metallurgical Engineers, Vol 
umes 83, 89, 93, 99, 104, in, 117, 122, 124, 
128, 133, 137 and 143, Institute of Metals 
Division. 



COPYRIGHT, 1941, BY THE 

AMERICAN INSTITUTE OF MINING AND METALLURGICAL ENGINEERS 

(INCORPORATED) 

PRINTED IN THE UNITED STATES OTT AMERICA 



THE MAPLE PRESS COMPANY, YORK, PA. 



FOREWORD 

DURING the fifteen years in which there has been a separate Institute of 
Metals Division volume of the TRANSACTIONS, there have thus been published 
just short of 400 technical papers comprising upward of 7000 pages. The 
present volume, containing 29 papers, is one of the largest of this series. How 
ever, quality has not been sacrificed for mere size, and the officers of the 
Division are proud of the caliber and scope of the papers here published. 

The annual lecture this year was given by Professor George Sachs, gen 
erally acknowledged as one of the world s outstanding physical metallurgists 
and one who contributed much to the literature of nonferrous metals. This 
lecture, on "Some Fundamentals of the Flow and Rupture of Metals," is 
an outstanding one, being somewhat unusual in that it held a strong appeal 
both to the theoretical and practical metallurgist. 

Possibly the membership at large does not realize how much the success 
of the programs and therefore of this volume depends on the Papers and 
Programs Committee, and this opportunity is taken to thank Dr. Cyril S. 
Smith, Chairman, Mr. E. E. Schumacher, Vice Chairman, and the other mem 
bers of that committee for the excellent work done. 

As Chairman of the Division, I also extend my sincere thanks to the 
other officers of the Division and Chairmen of the various committees, who 
have all cooperated toward a successful Division year. 

Thanks are also due the personnel of the Institute office, without whose 
help the Chairman would find himself most seriously handicapped. In this 
connection, it is felt that all of the members will regret to know that Mr. Louis 
Jordan, who for several years did so fine a job as Secretary of the Division, 
left in March to take a position with the O.P.M., in Washington, where no 
doubt his talents will be used to good advantage. The Division is fortunate, 
however, in obtaining the services of Mr. Frank T. Sisco to fill the office 
vacated by Mr. Jordan. 

D. K. CRAMPTON, Chairman, 

Institute of Metals Division, 1941. 
WATERBURY, CONN. 
June 14, 1941. 



$& 2 1 1941 



A.I.M.E. OFFICERS AND DIRECTORS 

For the year ending February 1942 

PRESIDENT AND DIRECTOR 
JOHN R. SUMAN, Houston, Texas 

PAST PRESIDENTS AND DIRECTORS 
DONALD B. GILLIES, Cleveland, Ohio 
H. G. MOULTON, New York, N. Y. 

TREASURER AND DIRECTOR 
KARL EILERS, Sea Cliff, N. Y. 

VICE-PRESIDENTS AND DIRECTORS 

WILLIAM B. HEROY, Houston, Texas W. M. PEIRCE, Palmerton, Pa. 

HENRY KRUMB, New York, N. Y. PAUL D. MERICA, New York, N. Y. 

ERLE V. DAVELER, New York, N. Y. LEROY SALSICH, Duluth, Minn. 

DIRECTORS 

JOHN M. BOUTWELL, Salt Lake City, Utah JAMES T. MAC KENZIE, Birmingham, Ala. 

HOLCOMBE J. BROWN, Boston, Mass. HARVEY S. MUDD, Los Angeles, Calif. 

CHARLES CAMSELL, Ottawa, Ont., Canada FRANCIS A. THOMSON, Butte 7 Mont. 

CHESTER A. FULTON, Baltimore, Md. J. R. VAN PELT, Jr., New York, N. Y. 

J. TERRY DUCE, San Francisco, Calif. H. Y. WALKER, New York, N. Y. 

C. A. GARNER, Hazleton, Pa. F. A. WARDLAW, JR., Inspiration, Ariz. 

H. T. HAMILTON, New York, N. Y. CLYDE E. WILLIAMS, Columbus, Ohio 

IRA B. JORALEMON, San Francisco, Calif. F. E. WORMSER, New York, N. Y. 

L. E. -YOUNG, Pittsburgh, Pa. 

SECRETARY 
A. B. PARSONS, New York, N. Y. 



DIVISION CHAIRMEN Acting as Advisers to the Board 

D. K. CRAMPTON (Institute of Metals), Waterbury, Conn. 

EUGENE A. STEPHENSON (Petroleum), Lawrence, Kansas 

CHARLES H. HERTY, JR. (Iron and Steel), Bethlehem, Pa. 

JULIAN E. TOBEY (Coal), Cincinnati, Ohio 

W. R. CHEDSEY (Education), Rolla, Mo, 

PAUL M. TYLER (Industrial Minerals), Washington, D. C. 

STAFF IN NEW YORK 

Assistant Secretaries Assistant to the Secretary 

EDWARD H. ROBIE E. J. KENNEDY, JR. 

CHESTER NARAMORE 

FRANK T. Sisco Business Manager, 

Assistant Treasurer "Mining and Metallurgy" 

H. A. MALONEY WHEELER SPACEMAN 



CONTENTS 

PAGE 

FOREWORD. By D. K. CRAMPTON 3 

A.I.M.E. OFFICERS AND DIRECTORS 4 

INSTITUTE OF METALS DIVISION OFFICERS AND COMMITTEES 7 

INSTITUTE OF METALS DIVISION ANNUAL AWARD CERTIFICATE 9 

INSTITUTE OF METALS DIVISION LECTURES AND LECTURERS 10 

PHOTOGRAPH OF GEORGE SACHS, INSTITUTE OF METALS DIVISION LECTURER 12 

PAPERS 

Some Fundamentals of the Flow and Rupture of Metals. By GEORGE SACHS. (Annual 

Lecture) (T.P. 1335) 13 

Internal Friction of Single Crystals of Copper and Zinc. By THOMAS A. READ. (T.P. 1309, 

with discussion) 30 

Time and Temperature Effects in the Deformation of Brass Crystals. By H. L. BURGHOFF 

and C. H. MATHEWSON. (T.P. 1288, with discussion) 45 

Flow of Solid Metals from the Standpoint of the Chemical-rate Theory. By WALTER KAUZ- 

MANN. (T.P. 1301, with discussion) 57 

Deformation and Recrystallization of Copper and Brass Hardness Microstructure and 

Texture Changes. By R. M. BRICK and M. A. WILLIAMSON. (T.P. 1299, with discussion) 84 
Corrosion of Copper and Alpha Brass Film-structure Studies. By J. H. HOLLOMON and 

JOHN WULFF. (T.P. 1311, with discussion) 93 

Some Practical Observations on Inverse Segregation. By DANIEL R. HULL. (T.P. 1287, with 

discussion) 98 

Self-diffusion of Silver. By WILLIAM A. JOHNSON. (T.P. 1272, with discussion) 107 

On the Equilibrium Solidification of Solid Solutions. (Abstract) By MORRIS COHEN and 

" WILLIAM?. KIMBALL. (T.P. 1256) 113 

Measurement of Irreversible Potentials as a Metallurgical Research Tool. By R. H. BROWN, 

W.L. FINK and M. S. HUNTER. (T.P. 1234, with discussion) 115 

X-ray Study of the Solid Solubility of Lead, Bismuth and Gold in Magnesium. By FRANK 

FOOTE and E. R. JETTE. (T.P. 1248, with discussion) 124 

Mechanism of Precipitation from the Solid Solution of Silver in Aluminum. By C. S. BARRETT, 

A. H. GEISLER and R. F. MEHL. (T.P. 1275, with discussion) *34 

Precision X-ray Study of the High-silver Aluminum-silver Alloys. By FRANK FOOTE and 

ERIC R. JETTE. (T.P. 1229) 151 

X-ray Analysis of Hot-galvanized Heat-treated Coatings. By F. R. MORRAL and E. P. 

Miller. (T.P. 1224, with discussion) 158 

Creep and Recrystallization of Lead. By ALBERT A. SMITH, JR. (T.P. 1227, with discussion) 165 
Tensile Properties of Rolled Magnesium Alloys Binary Alloys with Calcium, Cerium, 

Gallium, and Thorium. By John C. MCDONALD. (T.P. 1247, with discussion) 179 

Grain Orientation of Cast Poly cry stalline Zinc, Cadmium and Magnesium. By GERALD 

EDMUNDS. (T.P. 1244, with discussion) 183 

Studies upon the Corrosion of Tin Effects of Cations in Carbonate Solutions and Effects of 

Alloying Elements. By GERHARD DERGE and HAROLD MARKUS. (T.P. 1306) 198 

Effect of Composition on Physical and Chemical Properties of i4-karat Gold Alloys. By 

TRACY C. JARRETT. (T.P. 1249) 209 

S 



6 CONTENTS 

PAGE 
Beneficial Effects of Zirconium in Cast Nickel-silicon Bronzes. By F. R. HENSEL, E. I. LARSEN 

and A. S. DOTY. (T.P. 1237) 212 

Study of the Metallography and Certain Physical Properties of Some Alloys of Cobalt, 

Iron, and Titanium. By CHARLES R. AUSTIN and CARL H. SAMANS. (T.P. 1257, with 

discussion) 216 

Effect of Cold-work upon Electrical Conductivity of Copper Alloys. By D. K. CRAMPTON, 

H.L. BURGHOFF and J. T. STACY. (T.P. 1290, with discussion) 228 

Low-temperature Oxidation of Single Crystals of Copper. By BENJAMIN LUSTMAN and 

ROBERT F. MEHL: (T.P. 1317, with discussion) 246 

Preparation and Some Properties of High-purity Copper. By J. S. SMART, JR., A. A. SMITH, 

JR., and A. J. PHILLIPS. (T.P. 1289, with discussion) 272 

Solubility of Sulphur Dioxide in Molten Copper. By CARL F. FLOE and JOHN CHIPMAN. 

(T.P. 1308, with discussion) 287 

Solubility of Oxygen in High-purity Copper. By ARTHUR PHILLIPS and E. N. SKINNER, JR. 

(T.P. 1280, with discussion) 301 

Hydrogen Embrittlement of Pure Copper and of Dilute Copper Alloys by Alternate 

Oxidation and Reduction. By FREDERICK N. RHINES and WILLIAM A. ANDERSON. (T.P. 

1235, with discussion) 312 

Coalesced Copper Its History, Production and Characteristics. By H. H. STOUT. (T.P. 1238, 

with discussion) 326 

Coalescence Process for Producing Semifabricated Oxygen-free Copper. By JOHN TYSSOWSKI. 

(T.P. 1217) 335 

INDEX 341 

Contents of Volume 145 (Iron and Steel Division) 349 



INSTITUTE OF METALS DIVISION 



Established as a Division April 26, 1918 
(Bylaws published in the 1939 TRANSACTIONS Volume of the Division.) 

Officers and Committees for Year ending February 1942 

Chairman, D. K. CRAMPTQN, Waterbury, Conn, 
Past-chairman, EDMUND M. WISE, Bayonne, N. J. 
Vice-chairman, CYRIL S. SMITH, Waterbury, Conn. 
Vice-chairman, CARL E. SWARTZ, Cleveland, Ohio 
Treasurer, H. A. MALONEY, New York, N. Y. 
Secretary, FRANK T. Sisco, 29 West sgth St., New York, N. Y. 



W. M. CORSE, 1918-1920 
W. H. BASSETT, 1920-1922 
W. B. PRICE, 1922-1924 
G. K. ELLIOTT, 1924-1926 
P. D. MERICA, 1926-1928 
S. SKOWRONSKI, 1928-1930 



PAST CHAIRMEN 
ZAY JEFFRIES, 1930 
SAM TOUR, 1931 
C. H. MATHEWSON, 1932 
T. S. FULLER, 1933 
J. L. CHRISTIE, 1934 



Executive Committee 



1942 

FRANCES H. CLARK, New York, N. Y. 
W. L. FINK, New Kensington, Pa. 
J. T. NORTON, Cambridge, Mass, 



W. M. PEIRCE, 1935 

E. H. Dix, JR., 1936 

A. J. PHILLIPS, 1937 

R. F. MEHL, 1938 

R. H. LEACH, 1939 

EDMUND M. WISE, 1940 



1943 

ERIC R. JETTE, New York, N. Y. 

Louis W. KEMPF, Cleveland, Ohio 

GEORGE SACHS, Cleveland, Ohio 



1944 



J. R, FREEMAN, JR. 



W. H. BASSETT, JR. 
E. W. PALMER 



P. H. BRACE 
R. M. BRICK 

H. L. BURGHOFF 

ERIC R. JETTE 



C. C. BALKE 

F. E. CARTER 

G. H. CHAMBERS 



WILLIAM C. ELLIS, New York, N. Y. 
ALAN MORRIS, Bridgeport, Conn. 
KENT R. VAN HORN, Cleveland, Ohio 

Finance 

JOHN L. CHRISTIE, Chairman 
W. M. PEIRCE 
H. M. ST. JOHN 

Data Sheet 
LYALL ZICKRICK, Chairman 

F. N. RHINES 
S. SKOWRONSKI 

Papers and Programs 

E. E. SCHUMACHER, Chairman 

E. A. ANDERSON, Vice-chairman 

T. E. KlHLGREN 

ALAN MORRIS 
A. A. SMITH, JR. 
CYRIL S. SMITH 

Rare and Precious Metals 

ZAY JEFFRIES, Chairman 

W. P. SYKES, Vice-chairman 

G. M. HICKEY 
H. C. LAIRD 

H. OSBORG 

7 



JEROME STRAUSS 



KENT R. VAN HORN 
T. A. WRIGHT 



DANA W. SMITH 
W. P. SYKES 
T. A. WRIGHT 



A. J. PHILLIPS 

C. B. SAWYER 

E. M, WISE 



8 INSTITUTE OF METALS DIVISION 

Annual Lecture 

A, J. PHILLIPS, Chairman 

ZAY JEFFRIES R, H. LEACH H. S. RAWDON 

E. R. JETTE H. J, ROAST 

Annual Award 
C. S. BARRETT, Chairman 

D. K. CRAMPTON Louis W. KEMPF E. E. SCHUMACHER 

C. S. SMITH 

Mining and Metallurgy 
CARL E. SWARTZ, Chairman 

F. H. CLARK W. A. DEAN GEORGE M. HICKEY 
R, S. DEAN W. H. FINKELDEY S. L. HOYT 

SAM TOUR 

Membership 

J, T. KEMP, Chairman 

MORRIS COHEN, Vice-chairman 

W. H. BASSETT, JR. JAMES W. LAIST FRED P, PETERS 

A. ALLAN BATES R. W. E. LEITER B. E. SANDELL 

GERALD EDMUNDS W. A. MUDGE DANA W. SMITH 

F. R, HENSEL EARL W. PALMER FELIX E. WORMSER 

R. P. KOEHRING L. L. WYMAN 

Nominating 

CLAIR UPTHEGROVE, Chairman 

W. L. FINK T. S. FULLER J. T. NORTON 

N. B, PILLING 



ANNUAL AWARD CERTIFICATE OF THE INSTITUTE OF 
METALS DIVISION 

In 1933, the Institute of Metals Division of the American Institute of Mining and Metallurgi 
cal Engineers established its annual award of an engraved certificate to the author or authors of 
the paper that in the opinion of the award committee represents the most notable contribution to 
metallurgical science among the papers that have been accepted by the Division for presentation 
at one of its meetings and have been published by the Institute within the three years preceding 
the date of award. The award is made by the Division each February. There are no restrictions 
with respect to nationality, age or occupation of the author or authors. 

Awards have been made as follows: 

1934 Robert F. Mehl and Charles S. Barrett: Studies upon the Widmanstatten Structure, I 

Introduction. The Aluminum-silver System and the Copper-silicon System. TRANSAC 
TIONS (1931) 93, 78-110. 

1935 E. A. Anderson, M. L. Fuller, R. L. Wilcox and J. L, Rodda: The High-zinc Region of the 

Copper-zinc Phase Equilibrium Diagram. TRANSACTIONS (1934) in, 264-292. 

1936 Cyril S. Smith and W. EarlLindlief: A Micrographic Study of the Decomposition of the 

Beta Phase in the Copper-aluminum System. TRANSACTIONS (1933) 104, 69-105. 

1937 Arthur Phillips and R. M. Brick: Effect of Quenching Strains on Lattice Parameter and 

Hardness Values of High-purity Aluminum-copper Alloys. TRANSACTIONS (1934) m, 
94-112. 

1938 William L. Fink and Dana W. Smith: Age-hardening of Aluminum Alloys, I Aluminum- 

copper Alloy. TRANSACTIONS (1936) 122, 284-293. 

1939 Frederick N. Rhines and Robert F. Mehl: Rates of Diffusion in the Alpha Solid Solutions of 

Copper. TRANSACTIONS (1938) 128, 185-221. 

1940 Alden B. Greninger: Martensite Transformation in Beta Copper-aluminum Alloys. 

TRANSACTIONS (1939) 133? 204-221. 

1941 S. E. Maddigan and A. I. Blank: Recovery and Recrystallization in Long-time Annealing of 

70-30 Brass, TRANSACTIONS (1940) 137, 170-190. 



THE INSTITUTE OF METALS LECTURE 

An annual lectureship was established in 1921 by the Institute of Metals Division, which 
has come to be one of the important functions of the Annual Meeting of the Institute. In 1934 the 
Division established the custom of presenting a, certificate to each lecturer. 

A number of distinguished men from this country and abroad have served in this lectureship. 
The roll is quoted below: 

1922 Colloid Chemistry and Metallurgy. By Wilder D. Bancroft. 

1923 Solid Solution. By Walter Rosenhain. 

1924 The Trend in the Science of Metals. By Zay Jeffries. 

1925 Action of Hot Wall: a Factor of Fundamental Influence on the Rapid Corrosion of Water 

Tubes and Related to the Segregation in Hot Metals. By Carl Benedicks. 

1926 The Relation between Metallurgy and Atomic Structure. By Paul D. Foote, 

1927 Growth of Metallic Crystals. By Cecil H. Desch. 

1928 Twinning in Metals. By C. H. Mathewson. 

1929 The Passivity of Metals, and Its Relation to Problems of Corrosion. By Ulick R. Evans. 

1930 Hard Metal Carbides and Cemented Tungsten Carbide, By S, L. Hoyt. 

1931 X-ray Determination of Alloy Equilibrium Diagrams. By Arne Westgren. 

1932 The Age-hardening of Metals. By Paul D. Merica. 

1933 Present-day Problems in Theoretical Metallurgy. By Georg Masing. 

1934 Ferromagnetism in Metallic Crystals. By L. W. McKeehan. 

1935 Gases in Metals. By C. A. Edwards. 

1936 Diffusion in Solid Metals. By Robert F. Mehl. 

1937 Refractories. By R. S. Hutton. 

1938 The Nature of Metals as Shown by Their Properties under Pressure. By P. W. Bridgman. 

1939 The Creep of Metals. By D. Hanson. 

1940 Acceleration of Rate of Corrosion by High Constant Stresses. By Edgar H. Dix, Jr. 

1941 Some Fundamentals of the Flow and Rupture of Metals. By George Sachs. 




GEORGE SACHS 

Institute of Metals Division Lecturer, 1941 



Some Fundamentals of the Flow and Rupture of Metals 



BY GEORGE SACHS,* MEMBER A.I.M.E. " 

(Institute of Metals Division Lecturef) 



I deeply appreciate the honor of being 
selected to deliver the twentieth Annual 
Lecture of the Institute of Metals Division. 

The subject of my paper is extremely 
involved and voluminous, therefore I shall 
discuss only some more recent experimental 
material and speculate somewhat on the 
physical and mechanical fundamentals of 
these changes. 

The flow and the rupture of a solid 
material are rather different phenomena. 
It is therefore surprising that up to a rather 
advanced stage in metallurgical science 
these two fundamental phenomena have 
been simultaneously treated and forced 
into the general conception of the "failure" 
of solid material. Only in this century, and 
particularly during the last 20 years, have 
the very different physical meanings of 
flow and rupture been clearly recognized. 
Several hundred of the more recent publi 
cations deal with these subjects. 1 

LAW OF SIMILARITY 
The basic values commonly considered 
responsible for the mechanical behavior of 
any material are the stresses. From this 
standpoint both a large and a small article 
made from the same homogeneous metal 
and of geometrically similar shapes should 
behave identically, i.e., they should deform 
geometrically in a similar manner, if sub 
jected to the same stress values at corre 
sponding points. This "law of similarity" 
was introduced about 60 years ago simul- 

* Associate Professor, Metallurgical Department, 
Case School of Applied Science, Cleveland, Ohio. 

f Presented at the New York Meeting, February 
1941. Twentieth Annual Lecture. Manuscript received 
at the office of the Institute Jan. 24, 1941. Issued as 
T.P, 1335 in METALS TECHNOLOGY, June 1941. 

1 References are at the end of the paper. 



taneously by Barba 2 and Kick 3 and was 
derived from the classical theory of 
elasticity. 

As far as the plastic behavior of a metal 
is concerned, this rule appears to be valid. 
The conventional properties of large and 
small tension or compression test bars, 
machined from larger sections, are identi 
cal; while the force required to deform a 
section is proportional to its cross-sectional 
area and the work consumed in deforming 
geometrically similar articles in a geometri 
cally similar manner is proportional to its 
volume. The general trend of the stress- 
strain diagram of a metal is independent 
of the section size (Fig. i). No exceptions 
from this rule are known. 

However, if the deformation of a metal 
is continued until rupture occurs, and the 
breaking stress or the work consumed is 
measured, a test bar of large cross section 
appears to fail before or at lower stresses 
than does one having a small cross section. 
Thus, the specific impact energy has been 
often found to decrease with increasing 
size of the test specimen, particularly for 
materials with a limited ductility. 4 Another 
exception from the rule of similarity is 
offered by fatigue-strength values ; which 
usually decrease with increasing section 
size. 5 - 6 

The conclusion must be drawn from these 
experimental facts that the ductility of a 
homogeneous metal decreases with increas 
ing section size (Fig. i). This effect has 
been actually confirmed by Docherty, 7 who 
observed that in static notch bending tests 
large specimens failed earlier and in a more 
brittle fashion than small specimens ma- 



SOME FUNDAMENTALS OF THE FLOW AND RUPTURE OF METALS 



chined from the same section (Fig. 2). It 
also appears to be common commercial 
experience that small sections can be cold- 
reduced much further than large sections. 
It may be mentioned in this connection 
that none of the conventional mechanical 




CROSS SEZT/OK 



STRAIN >- 

FIG. i. DIAGRAMMATIC REPRESENTATION OF 

STRESS-STRAIN DIAGRAMS FOR A LARGE AND A 
SMALL CROSS SECTION. 

tests give any measure for the ductility 
corresponding to the behavior of the metal 
in working or in service. Certainly, the 
elongation in a tensile test has little rela 
tion to the actual ductility; and the con 
traction in area at the rupture of a 
tensile test may indicate losses of ductility 
in some cases, but completely fails to do so 
in others. For example, the very ductile 
annealed duralumin has about the same 
contraction in area as the much brittler 
heat-treated condition. Also, a steel heat- 
treated to a high hardness frequently shows 
a higher contraction in area than it does 
in a softer condition, which is contrary to 
that indicated under actual performance. 
It is also known that small inclusions, such 
as the oxides in steel and copper, 8 con 
siderably reduce the contraction in area in 
tension tests, but have comparatively 
little effect on the general ductility. Other 
tests, such as the various bending or form 
ing tests, have not been sufficiently explored 
with respect to ductility. In fact, little has 
been done up to the present time to analyze 
this fundamental property, while a large 
amount of work is being expended con 



tinuously in the investigation of various 
commercial forming tests. 9 - 10 - 11 

THEORIES OF PLASTICITY 

The next problem regarding the flow and 
rupture of a metal is that of the effect of 



*M 



45 

xrr 



_ 
A 

T 



01 L 34 

51ZE &" OF 5BUf{R , /A/rvy 

FIG. 2. EFFECT OF SECTION SIZE ON DUCTIL 
ITY OF NOTCHED MILD STEEL BARS SUBJECTED 
TO BENDING (DOCHERTY). 

various types of stresses, such as tension, 
compression and shear, and of combina 
tions of these simple types of stresses. 




FIG. 3. DIAGRAMMATIC REPRESENTATION OF 
SHEAR STRESSES ACTING ON DIFFERENT PLANES 
OF STRESSED ROD. 

As far as the flow of the metal is con 
cerned, it was found in an early stage of 



GEORGE SACHS 



scientific metallurgy 1 that the principal 
determining factor is the maximum value 
of the shear stresses that are created in a 
strained metal (Fig. 3). If a metal is sub- 




FIG. 4. RELATION OF MAXIMUM SHEAR STRESS 
TO PRINCIPAL STRESSES. 

jected to any combination of homogeneous 
stresses, shear stresses are created in any 
section, on any imaginary set of planes. In 
one particular plane this shear stress, / max 
will be larger than in any other plane. 
According to the " maximum shear stress" 
hypothesis, a metal will flow under any 
combination of stresses (or " stress state") 
if the maximum shear stress reaches a 
critical value. 

While this conception at first appears 
very lucid and reasonable, a simple con 
clusion reveals that it is incompatible with 
the facts. According to general experience, 
a strained metal can progressively assume 
any desired cross-sectional shape; the 
dimensions can Increase or decrease, 
depending upon the particular combination 
of applied stresses and determined by the 
shape of the tools. However, in a strained 
metal one and only one set of planes will 
usually exist where the shear stress is higher 
than that on any other set (Fig. 4). This 
set is, according to the laws of elasticity, 
the set of planes that are inclined at an 
angle of 45 to both the directions of the 



largest and the smallest normal or principal 
stress (designated s\ and $3, respectively). 
The maximum shear-stress law postulates 
that this shear stress determines the plastic 







FIG. 5. DIAGRAMMATIC REPRESENTATION OF 

CHANGE OF GENERAL SHAPE OF ROB IN WHICH 
ONLY ONE SYSTEM OF SLIP PLANES ACTS. 

flow, which therefore will occur only in the 
direction of the shear stress. This, however, 
creates a rather restricted and special type 
of deformation (Fig. 5). A round section 
will be converted into an elliptical one, 
without any change of the one transverse 
dimension. And the change of shape will 
not be affected by any stress acting in this 
transverse direction, as long as. this stress 
does not become predominant. 

An alternative conception was intro 
duced as early as 1870 by the French 
scientists St. Venant 12 and Levy, 13 who 
compared the flow of solid metals to that of 
liquids. A liquid will flow in all directions, 
the extent of these strains depending upon 
the ratios of the corresponding stresses and 
changing continuously with the stresses. 
Such a continuous relation between stresses 
and strains (Fig. 6) has been confirmed in 
recent investigations. 14 Any modern theory 



i6 



SOME FUNDAMENTALS OF THE FLOW AND RUPTURE OF METALS 



of plasticity must be based on the assump 
tion of such a continuous relation. 

While mathematicians have proposed 
numerous more or less complicated combi- 



tubing, which is simultaneously subjected 

to a controlled internal pressure (Fig. 7). 

The other experimental fact, which will 

be discussed later in a more detailed 



1.0 



" 0.4 
** 0.2 



s - 



^ -a,6\ 



-23 




























































/ 


ft 


















/ 


















? 














J 


ft 


^ 












1 


^ 


^ 


/ 














/ 



















J 


t 
















9 J 


f* 


















t 









































-10 0.8 0.6 #4. &Z-O+0.Z 44 Q.6 O.Q l.ff 



FIG. 6. RELATION BETWEEN STRESSES AND STRAINS IN PLASTIC FLOW OF SOME METALS (LODE). 



nations of stresses as the fundamental factor 
responsible for the flow of metals, extensive 
experimental work has revealed two impor 
tant facts. 14 15 

First, it has been found that when vary 
ing the stresses in any possible manner the 
maximum shear stress during flow will vary 
with the stress combination, but only 
between the yield strength in tension or 
compression and a value approximately 
10 to 15 per cent higher than the yield 
strength (Fig. 7). This highest value of 
flow resistance will be observed in any 
strain condition where one dimension is 
not changed, as in rolling. This strain state 
also occurs in torsion. 

Another stress state of this type, which 
can be experimentally investigated in an 
exact manner, is realized in a tube that is 
being subjected to tension in two directions, 
longitudinally and circumferentially, the 
one tension being twice as high as the 
other (.?! = 2^2). The experimental proce 
dure consists of tension tests on thin-walled 



manner, is that the addition of hydrostatic 
pressure affects only slightly the flow stress. 
This variation can be neglected as far as 
the stresses encountered under commercial 
conditions are concerned. 

The theory that quantitatively agrees 
with the experimental facts is the so-called 
"strain-energy theory" advanced by v. 
Mises 16 and Haigh. 17 This conception con 
siders the totality of the shear stresses or 
the sum of the squares of the shear stresses 
responsible for the plastic flow, according 
to the following formula: 



where t\, t%, t$ indicate the three principal 
shear stresses, Si, $2, ss the three principal 
normal stresses and the yield strength in 
tension (or compression). This theory has 
attracted considerable attention and ap 
pears to represent the closest approach to 
experience that can be treated mathe 
matically in a fairly simple manner. Fre- 



GEORGE SACHS 



quently, however, the simpler maximum 
shear-stress theory will be sufficiently 
accurate for practical purposes. 




;y^ 



*" 



j^CiRCUf 



fLDk 



r/i/ry 



5TRE55 



FIG. 7. EFFECT OF BIAXIAL TENSION ON FLOW 

STRESS AND DUCTILITY OF STEEL TUBES. 




FIG. 8. SINGLE BRASS CRYSTAL, STRETCHED IN 
TENSION TO BREAKING POINT. 

On the other hand, no lucid physical 
conception is associated with the shear- 
strain theory. Attempts have been made, 
therefore, by Sachs 17 and, more completely, 
by Cox and Sopwith 18 to develop a condi 
tion of plasticity from the fundamental 
conceptions of plastic flow of single 
crystals. Single crystals usually are rod or 
plate-shaped samples consisting of only 



one crystal grain (Fig. 8). Such a single 
crystal flows if the shear component (Fig. 
9) on a crystallographic plane, the slip 




FIG. 9. DIAGRAMMATIC REPRESENTATION OF 

SLIP SYSTEM OF SINGLE CRYSTAL. 
N, normal to slip plane. 5, slip direction. 

plane, in a crystallographic direction, the 
slip direction, reaches a critical value. The 
principal stresses required to produce 
plastic flow in a single crystal therefore 



FIG. 10. DIAGRAMMATIC REPRESENTATON 
OF GROUP OF SINGLE CRYSTALS STRAINED IN 

TENSION. 

depend upon the angles between the slip- 
plane direction and the directions of stress, 
these yield-strength values usually varying 



i8 



SOME FUNDAMENTALS OE THE FLOW AND RUPTURE OF METALS 



considerably for differently oriented crys 
tals (Fig. 15). The total average of this 
flow stress for a group of crystals, con 
sisting of all possible orientations (Fig. 10), 



&o 






y 



B/AXIAL 




FIG. ii. DIAGRAMMATIC REPRESENTATION 

OF FRACTURES CAUSED BY VARIOUS STRESS 
STATES. 

can be numerically calculated for various 
stress states. It has been found that the 
average maximum shear stress will differ 
by about 15 per cent for the two extreme 
conditions of pure tension and of biaxial 
tension with the one tension twice the 
value of the other. These rather tedious 
calculations show that fundamentally the 
flow of a homogeneous crystalline aggregate 
is determined by crystallographic phe 
nomena. Unfortunately, however, the crys 
tallographic theory of plastic flow has not 
been sufficiently advanced to be expressed 
by simple equations, which can be readily 
used for calculations. 

There exists one rather unsatisfactory 
exception to the laws that determine the 
beginning of plastic flow. This is annealed 
low-carbon steel, which usually flows when 
the maximum shear stress reaches a con 
stant value, not dependent upon the other 
components of the stress state. 14 This rela 
tion also appears to be associated with the 
presence of the familiar yield-point elonga 
tion or jog in the stress-strain diagram. At 
the present time, these phenomena have 



not been adequately explained. While the 
conception of a brittle grain-bound ry 
skeleton in the soft steel 20 explains some 
of the features of the yield point, it does 
not permit any positive conclusion regard 
ing the conditions required for the begin 
ning of plastic flow. 

Apparently there is another fundamental 
conception that does not apply to soft 
steel. As previously discussed, it is usually 
assumed that a metal will flow locally if 
the stresses in a small volume correspond 
to the plasticity conditions. No deviations 
from this law have been observed for most 
metals. Soft steel, however, according to 
some accurate measurements 21 will with 
stand in localized areas, such as around a 
hole in a stressed beam, stresses more than 
2.5 times the expected values without 
plastic flow. Apparently soft steel behaves 
differently under uniform than under non- 
uniform stress conditions where high stress 
gradients exist. 

EFFECT OF HYDROSTATIC PRESSURES 
ON PLASTIC FLOW 

Large variations in the stress state can 
be realized by carrying out experiments in 
a high-pressure chamber. This has been 
done particularly by Bridgman 21 and 
numerous unusual changes of physical 
properties have been found to occur with 
increasing pressures. These pressures are of 
the hydrostatic type or of the same magni 
tude in all directions. 

The resistance to plastic flow is also 
somewhat increased by the addition of 
hydrostatic pressure. The rate of increase 
with increasing pressure, however, is com 
paratively small and can be neglected as 
far as the commercial behavior of the 
metals is concerned, as previously men 
tioned. This is expressed by the relation 
that the maximum shear stresses are not 
affected by the addition or subtraction of a 
hydrostatic pressure ( p): 

/max = (Si - p) - (Si - p) = Si - S 3 



GEORGE SACHS 



It has been observed also that hydro 
static pressures do not affect the strain 
conditions or type of plastic flow in metals. 
This may result in somewhat unusual 



n). Thus, biaxial pressure results in a 
tension-type failure, while, correspond 
ingly, biaxial tension produces a compres 
sion-type strain. 




FIG. 12. FRACTURES OP CAST STEEL SUBJECTED TO PURE TENSION AND TO BIAXIAL COMPRESSION 

(Ros AND ETCHINGER). 



f\l 

ClRCUMFERNT. 
UIMPRESSION 


o *| U- 




*V^ _^f~~ti. 


O 1 


I 


1 LONGITUDINAL % 


81 


-h h 


i 


"~*~ i 


NO RHO.IRL 


x^ 






5TRRIN 


TZvJjx 


& 


1 


C 
UJNGITUlUNfU. 
STRETCHING 


CUL 


\ 






H^ 


1 


L J 




^- ^*<t^////y///w/////w/////^^^ 





NO tlfKUMF. 

STRAIN 



-rry 

1 - r 



iOf/0/Tuaff/ftt. 



FIG. 13. DIAGRAMMATIC REPRESENTATION OF VARYING STRAIN CONDITIONS IN A DEEP-DRAWN 

SHELL. 
Small arrows indicate stresses acting at respective points. 



phenomena, as reported by Bridgman. 22 If 
a cylindrical bar is passed through a con 
tainer and subjected to a high biaxial 
pressure (/ 2 = s s) this bar fails (Fig. n) 
in fundamentally the same manner as a 
regular tension test bar (Fig. 12), dis 
regarding the fact that only compression 
stresses are present. This is explained by 
the mechanical relation that biaxial pres 
sure is identical with tension plus a 
superimposed hydrostatic pressure of the 
same magnitude as the tension stress (Fig. 



Generally, the same strain condition can 
be created by an infinite number of stress 
states, which, however, differ only by the 
hydrostatic pressure applied. 

This is confirmed also by the observation 
that the alignment of the crystal grains in 
a specific metal depends only upon the 
general change of shape or strain condi 
tions. 24 - 25 Thus various processes, such as 
tension, drawing, rolling or swaging a rod 
to the same final shape result in the same 
fiber structure. A particularly instructive 



2O 



SOME FUNDAMENTALS OF THE PLOW AND RUPTURE OF METALS 



illustration is offered by a drawn cup in 
which, according to Herrmann and Sachs, 25 
virtually all possible strain conditions are 
realized (Fig. 13). The compression-type 
strain A occurs at the edge of the cup, 
which is subjected mainly to circum 
ferential compression Ai, and also in the 
bottom A 2 where biaxial tension acts. A 
condition similar to rolling, in which one 
dimension is not changed (B) is also found 
at two points, somewhere in the upper part 
of the wall (Bi) where the original thick 
ness of the blank is preserved and at the 
corners (J5 2 ) where the fibers are mainly 
stretched but the diameter of the blank is 
retained. Near the latter point, a tension- 
type strain and stress (C) occur. The 
X-ray analysis of the different points has 
revealed only one series of fiber structures, 
these being determined by the respective 
strain conditions and not by the widely 
varying stress conditions. 

EFFECT OF STRESS STATE ON DUCTILITY 

The rupture of a metal, on the contrary, 
is decidedly affected by the stress state, 
and particularly by the superposition of 
hydrostatic stresses. 

While a general theory of the rupture of 
metals has not yet been developed, it is 
apparent that the majority of service 
failures can be attributed to the presence 
of excessive tension stresses. 26 

Therefore, the superposition of a hydro 
static pressure, which reduces the tension, 
also diminishes the tendency to rupture and 
increases the ductility. This has been 
particularly observed with rocks that 
usually are brittle but become increas 
ingly plastic with increasing hydrostatic 
pressures. 27 28 

This effect is less pronounced with metals 
which are, in most cases, fairly ductile in 
ordinary atmospheric pressures. However, 
the ductility of any metal is also con 
siderably increased by the addition of 
hydrostatic pressures, 14 as illustrated in 
Fig. 12 for the contraction in area of a cast 



steel, and in Fig. 14 for the stress-strain 
diagrams of an aluminum bronze in tension 
and cast iron in compression. 

The beneficial effect of hydrostatic pres 
sures is being widely utilized in the manu 
facture and fabrication of the metals. 
There are two basic methods of creating 
high pressures. The first is the use of 
closed passes or dies, as in die forging, rod 
rolling or extruding, which force the metal 
to flow through narrow orifices. The second 
is the presence of high friction forces, which 
will be discussed later in detail. The objec 
tive is to eliminate tension stresses in 
order to reduce the tendency for brittle 
fractures. 26 

Another aspect of this problem is 
equally important. Any superimposition of 
hydrostatic tension should increase the 
tendency to rupture and consequently 
reduce the ductility. In the extreme case of 
high triaxial tension, any material should 
behave in a completely brittle manner. 
While such a condition may be approached 
in practice, as in quenching or welding, it 
can be experimentally realized and followed 
up only to a very limited extent. The most 
lucid experimental evidence of an embrit 
tling effect of multiaxial tension has been 
again observed in tubing that is simul 
taneously subjected to longitudinal and 
circumferential tension. 29 - 30 By increasing 
the latter to a value where it becomes 50 
per cent of the longitudinal tension, the 
ductility, measured by the local contraction 
or elongation at the failure is reduced 
materially (Fig. 7). Thus an additional 
transverse tension has a pronounced em 
brittling effect. 

The presence of such transverse tension 
also accounts for the much discussed effects 
of sharp notches on the tensile properties 
of the metals. 31 32 33 Practical experience 
evidently indicates that external or internal 
notches are a frequent and important 
source of service failures of the brittle type. 
However, the results so far published on 
this subject disagree with the theoretical 



GEORGE SACHS 



21 



conceptions 34 therefore a discussion of this 
subject must be postponed until new 
experimental data are available. 



by Schmid 36 and the amount of movement 
in this direction or shear strain, suggested 
by Sachs. 37 With these units, the stress- 



r 

^ 






Ci 


75 T JAO 


N 




*. 


~ * 


~^~ 





---+ 


r 


^ ** 


.-__ 




/rm 


mess. 


^ 


tiwtMi 


" a *^ 







JQ 



ZO 



flEOVCTif/ <w 

I 4 -EFFECT OF HYDROSTATIC PRESSURE ON DUCTILITY OP CAST IRON (COMPRESSION) AND 
14- ****<- ALUMINUM BRONZE (TENSION) (ROS AND ElCHINGER). 



\" 



o 



W 



paLY&iysrtiUJHf 



20 
5t/ttH 



t 

FIG 15 -STRESS-STRAIN DIAGRAMS OF SINGLE ALUMINUM CRYSTALS IN 

AND USING MECHANICAL AND CRY STALLO GRAPHIC COORDINATES, RESPECTIVELY 



STRESS AND STRAIN or SINGLE CRYSTALS 

AND OF POLYCRYSTALLINE AGGREGATES 

Plasticity is a basic property of metallic 
crystals. During the last 20 years the 
mechanical properties of single crystals 
have been studied extensively 35 36 and it 
has been found that differently oriented 
single crystals of a specific metal usually 
show widely varying stress-strain diagrams 
for any type of testing, such as tension, 
compression or torsion (Fig. 15). These 
diagrams can be converted into a single 
characteristic curve by replacing the con 
ventional mechanical stress and strain 
units, such as tension and reduction in 
area (or elongation) by corresponding 
crystallographic units. These crystallo- 
graphic units are the shear-stress ^com 
ponent on the slip plane in the direction of 
slip, proposed by Taylor and Elam 35 and 



strain diagrams for differently oriented 
crystals of aluminum 38 and zinc 36 are 
almost identical (Fig. 15), and^the same 
stress-strain curve is obtained in tension 
and in compression. 39 

The yield strength, the strain-hardening 
and correspondingly the ultimate strength 
of a ductile polycrystalline aggregate is 
always considerably higher than the aver 
age strength of the single crystals that 
constitute the aggregate (Fig. 15)- To 
explain this difference, many metallurgists 
have accepted Beilby s hypothesis of very 
hard and strong amorphous layers in the 
grain boundaries of the individual crystal 
grains. However, any attempt to prove the 
existence of such amorphous metal has 
been unsuccessful. 

On the contrary, Chalmers 40 found that 
the presence of a longitudinal grain bound 
ary in "bicrystal" tin specimens increases 



22 



SOME FUNDAMENTALS Of THE FLOW AND RUPTURE OF METALS 



the yield strength only, if the two crystal 
grains differ in their orientation, and that 
the increase is approximately proportional 
to this difference in orientation (Fig. 16). 



r 
1 

I 
I, 

*\ 

I s - 

1 


















S 












^ 




^ 


/ 

\1t 


n 








^ 


s< 






f" 

*r 

M 






/ 


*s 








wf- 


*m 


fS 













FIG. 1 6. YIELD STRENGTH OF TIN SPECI 
MENS CONSISTING OF TWO CRYSTALS WITH 
LONGITUDINAL BOUNDARY (CHALMERS). 

Thus, the difference in strength between 
single crystals and a polycrystalline aggre 
gate can be explained by the fact that 
adjacent crystals mutually obstruct their 
crystallographic deformation. While a 
single crystal becomes oval in tension (Fig. 
8), the crystal grains in an aggregate are 
being forced to alter their shape to corre 
spond to the general change of shape of the 
strained article. According to Taylor 41 this 
results in complex slip movements within 
the crystals accompanied by an increased 
strain-hardening. If differently oriented 
crystals are forced into a given shape, five 
different gliding systems become active. 
The extent of the shear movement occur 
ring on each gliding system and the sum of 
these shears on various gliding systems can 
be calculated, and the corresponding strain- 
hardening can be determined from the 
stress-strain diagram of a single crystal. 
The stress-strain diagram of a polycrys 
talline aluminum calculated on this basis 
agrees with the experimentally observed 
curve (Fig. 17). So far, this tedious calcula 
tion has been carried out only for tension. 
This superposition of various slip move 
ments also explains the preferred orienta 



tion that occurs on cold-working. In a 
single crystal, any strain will cause the 
crystal lattice to rotate in such a manner 
that some important crystal directions 
approach the directions of strain. Funda 
mentally, the same reorientation process 
takes place in the grains of a polycrystalline 
metal. This " fibering," however, is more 
complicated than and different from that 
of single crystals. According to Taylor, 41 
the changes in orientation of the individual 
crystal grains can be calculated from the 
slip movements required to alter the 
external shape during deformation. While 
this theory offers an approach to the solu 
tion of the extremely complex problem of 
crystallographic fibering during working, 
investigations by Barrett 42 indicate that 
it needs further refinement in order to 
satisfactorily explain the experimental 
evidence. 

The crystallographic fibering also deter 
mines to a certain extent the physical and 
mechanical properties in worked and in 
annealed metals (Fig. 18). Some metal 
products are commercially produced with 
such a pronounced preferred orientation 
that their structures and properties ap 
proach those of a single crystal. Thus, in 
annealed copper sheet, all crystal grains 
may be, within certain limits, aligned with 
the axes of their cubic crystal lattice 
parallel to the rolling, transverse and nor 
mal directions, 36 - 43 - 44 and such sheet with 
a " cubic structure" has directional prop- 
perties almost corresponding to those of a 
single crystal (Fig. 18); i.e., specimens 
taken in different directions have very 
different properties. The agreement be 
tween the properties of such sheet and 
those of a similarly oriented single crystal 
is particularly good with respect to the 
elastic properties. 44 The same is true for the 
magnetic qualities of iron -nickel alloy 
sheet 45 and iron-silicon alloy sheet, the 
latter being commercially produced with 
a highly preferred orientation and direc 
tionality of the magnetic properties. 46 - 47 



GEORGE SACHS 



2 3 



While considerable progress has been not apply when plastic flow precedes the 

made during recent years regarding the failure; and this deformation introduces 

fundamental conceptions of the plastic complicated and as yet not explained 

flow in single crystals and polycrystalline phenomena. Single crystals of the more 



** 






in 



10 



iff 



30 



BTftl. 
GXE&, 



40 



SING& ex* 




10/7 



FIG. 17. TENSION-EXTENSION CURVE OF POLYCRYSTALLINE ALUMINUM, DERIVED PROM SHEAR 
STRESS-SHEAR CURVE OF SINGLE CRYSTAL (TAYLOR). 




FIG. 1 8. DIRECTIONALITY or VARIOUS PROPERTIES OF ANNEALED COPPER SHEET WITH "CUBIC" 

FIBER STRUCTURE. 

Cubic axes parallel to rolling, transverse and normal directions. 



aggregates, much less success has been 
achieved with the conditions of rupture. 
It appears^ from extensive experiments by 
Schmid 36 on single crystals of the brittle 
type, that cleavage occurs if the normal 
stress on the crystal plane that functions 
as the cleavage plane attains a critical 
value. However, this simple relation does 



ductile metals do not develop cleavage 
fractures but fail in tension by a process 
similar to the necking of the polycrystalline 
metals (Fig. 8). At the present time the 
factors responsible for this type of failure 
and the conditions that govern the failure 
of metals in commercial working are not 
known. 



24 SOME FUNDAMENTALS OE THE FLOW AND RUPTURE OF METALS 



The following parts of this lecture will, 
for this reason, deal only with the principal 
factors that influence the plastic flow in 
commercial working of metals. 



curve describes the resistance of a metal to 
any type of strain such as tension, com 
pression, shear or combinations of them. 
It is, also, possible to derive from such crys- 




FIG. 19. STRAIN-HARDENING OF COPPER IN TENSION, COMPRESSION AND TORSION (Ros AND 

EICHINGER). 

tallographic data a corresponding stress- 
strain curve describing the behavior of a 
polycrystalline aggregate subjected to any 
type of strain. 

This practically important problem 
attracted the interest of the metallurgist 
in an early stage in the development of 
metallurgical science. Ludwik 49 proposed 
that the extent of shear along the con 
tinuously changing plane of maximum 
shear stress determines the magnitude of 
strain-hardening. Thus, the stretching of a 
certain initial length (/ ) to a final length 
(Zi) will produce the same strain-hardening 
as the compression of an initial height (/i) 
to a height (Zo). Ludwik also introduced a 
logarithmic strain function, the "effective 
strain": 



STRAIN-HARDENING OF METALS 

One of the unsolved mysteries of metal 
lurgy is the source of the strain-hardening 
occurring during cold-working. This im 
mensely important property, which is 
characteristic not only of metals but also 
of other crystalline matter such as rock- 
salt, 48 has not yet been given a convincing 
theoretical explanation. Many metallur 
gists fear that the tools available are not 
sensitive enough to discover the source of 
the strain-hardening and that the most 
modern and unfortunately least lucid 
physical conceptions will be required to 
solve this problem. 

As previously discussed, the capacity for 
strain-hardening is an inherent property of 
any single metallic crystal (Fig. 15). It is 
well known that strain-hardening is re 
tained only at comparatively low tempera* 
tures. Heating eliminates this effect and 
approximately reproduces the original 
softness. 

It has been also mentioned that the 
strain-hardening of single crystals of a 
specific metal can be represented by a 
characteristic stress-strain curve. - This 



h h 

= log. - = - log, ~ 

/o n 



which represents the shear strain of a 
polycrystalline material. The stress-strain 
diagrams 15 of a metal both in tension and 
in compression plotted with the effective 
strain as abscissa are identical (Fig. 9), 
and the same is true for the stress-strain 



GEORGE SACHS 



diagram in torsion (or shear) if it is con 
sidered that the flow stress is 10 to 15 per 
cent higher in torsion than in tension or 
compression, as previously discussed (Fig. 

7). 

The commercial working processes may 
be either of the tension, compression or 
shear type, or of a combination type. As 
the volume in plastic deformation remains 
virtually constant, one principal strain is 
either tension or compression, while the 
other two have the opposite sign. The 
strain-hardening is determined by the strain 
that is different from the other two with 
regard to its sign, such as the stretching 
in drawing or the reduction of thickness in 
rolling. 

While these relations offer at least simple 
approximations, which may be useful in 
practice, attempts have been made also 
to develop strain functions that closely 
agree with the experimental results. Among 
these functions, the so-called " octahedral 
shear" developed by Nadai 50 has attracted 
particular interest. However, as previously 
mentioned, a correlation of the strain- 
hardening of polycrystalline aggregates 
with that of single crystals has not yet been 
achieved in a general manner. 

EFFECTS OF TOOL CONTACT AND FRICTION 
ON PLASTIC FLOW 

Any attempt to apply the described 
fundamental conceptions to plastic flow 
encounters the principal difficulty that the 
flow in a commercial working process is not 
uniform. 51 

This discussion will not consider any 
chemical source of nonuniformity, such as 
that represented by segregations or "den- 
drites," but assumes that the metal 
subjected to the working process is homo 
geneous. Nevertheless, it is usually observed 
that various parts of an article have differ 
ent properties, particularly the surface and 
the core. This may be caused by various 
factors such as: 

i . The localized action of the tools. 



2 . The presence of friction between metal 
and tool. 

3. The heat developed by deformation 
and friction. 

4. Impacts associated with high-speed 
working. 



CQNTfiCT LEHGTH 



tltttttt 



WITH FRICTION 



\ 



FIG. 20. DIAGRAMMATIC REPRESENTATION 
OF EFFECTS OF CONTACT LENGTH AND FRICTION 
IN FORGING. 

Little work has been done so far toward 
analyzing working processes in these 
respects, and the following remarks are 
therefore speculative. 

The effect of the tools in working a metal 
depends particularly upon the size of the 
working surface of "area of contact. " 
While it is impossible as yet to develop 
quantitative relations, it appears that the 
ratio (l/h) between the length of contact 
(I) in the principal direction of flow and the 
thickness (ti) of the metal is the deter 
mining factor. A thin bar will flow uni 
formly under the action of large dies (Fig. 
20), developing mainly compression stresses 
in the direction perpendicular to the die 
surface. However, a bar subjected to the 
localized attack of narrow tools (Fig. 20) 
will be plastically deformed only in the 
vicinity of the contact area. A load con 
centrated at a single point creates a very 



26 



SOME FUNDAMENTALS OF THE FLOW AND RUPTURE OF METALS 



nonuniform stress distribution and com 
pression or tension may be present at a 
specific point in both the axial and the 




SSI 



FIG. 21. EFFECTS OF CONTACT LENGTH IN 

DRAWING ON PROPERTIES OF PRODUCT. 

a, drawing with large die angle. 

5, drawing with small die angle. 

c, tension (stretching). 

transverse directions. Thus, with a given 
metal thickness, the deviations from uni 
formity will increase with decreasing con 
tact length. This conclusion has been 
confirmed for any type of working process, 
and has been thoroughly examined for 
wire drawing. 52 A wire drawn through a 
wide-open die (Fig. 21) having a short 
contact length will develop a nonuniform 
stress and strain distribution. The strength 
of the surface layers will be considerably 
higher than that of the core, and high 
residual stresses will be retained, which 
will cause considerable distortion during 
machining. On the contrary, a wire drawn 
with an acute-angle die will be in every 
respect more uniform, approaching the 
condition of a uniformly stretched tensile- 
test bar. 

The effects of friction on plastic flow also 
primarily depend upon the ratio of contact 



length to metal thickness (Fig. 20). With 
a short contact length little friction is 
present and its effect mainly consists of 
accentuating the nonuniformity of the 
stress and strain states. With a large 
contact length, however, two types of fric 
tion effects are developed. First, the friction 
acting along the contact length and oppos 
ing the metal flow may reach a high force. 
With very thin metal, as in rolling thin 
strip or die forging with a thin flash, the 
friction may completely prevent any flow, 
and the equipment will fail before the 
metal is displaced. Under less extreme con 
ditions, a high transverse pressure is 
created, and consequently a high pressure 
is required to deform the metal. This, 
however, corresponds to the superposition 
of hydrostatic pressure and results in the 
beneficial effect of increasing the ductility, 
as previously discussed. This second effect 
is valuable in working sensitive metals. 

SPEED AND IMPACT EFFECTS 

One of the least analyzed and explained 
factors in mechanical working is the work 
ing speed. It has been recognized that the 
following effects can be influenced by the 
velocity of deformation: (i) the properties 
of the metal, (2) the impact effect, or 
stress and strain concentration in the 
metal adjacent to the tool, (3) the heat 
development and dissipation. 

The metal properties depend only 
slightly upon the working speed in cold- 
working, while in hot-working the flow 
resistance of the metal increases consider 
ably with increased working speed. The 
displacement of the cold-working range to 
higher temperatures by an increased rate 
of working 51 - 53 is of practical importance 
and results in the restriction of the range 
of hot-working temperatures. This explains 
the benefit derived from the slow working 
of many alloys, such as high-alloyed steels 
and magnesium alloys. The peculiar and 
unexplained brittleness present in some 
alloys, for instance, the alpha brasses, is 



GEORGE SACHS 



found at different temperatures, depending 
upon the rate of strain. 54 Another unusual 
phenomenon is the increase of the yield 
point jog and of the associated tendency of 




FIG. 22. DIAGRAMMATIC REPRESENTATION 
OF EFFECT OF IMPACT ON STRAIN AND STRESS IN 
UPSETTING. 

stretcher strain formation by increasing 
the speed of cold-working low-carbon 
steels. 65 

In commercial practice it is widely 
recognized that it is extremely important 
to overcome the impact effects. 51 The 
impact appears to be the primary factor 
limiting the working speed in many com 
mercial processes. 

The impact effect may be fundamentally 
denned as the additional stresses and 
strains resulting from deviations from static 
equilibrium. A body is in static equilibrium 
when it is at rest or when it is moving with 
uniform velocity. Accelerations and de 
celerations, however, such as those particu 
larly present at the beginning of a working 
process, will result in stresses in addition 
to those required to overcome the flow 
resistance of the metal. This impact effect 
is more important in the tension-type than 
in the compression- type working processes. 
Generally, the acting force must be trans 



ferred through the metal to the various 
sections being worked. In compression-type 
working, such as is represented by the 
upsetting of a blank (Fig. 22), the main 



-TOOLS WTHSHftRP LOfiNEAS 

* flGMOEQ Off TOOLS 




FIG. 23. DIAGRAMMATIC REPRESENTATION 
OF EFFECTS OF TOOL DESIGN AND SPEED IN 
WORKING ON FORCE-STROKE DIAGRAM IN DEEP 

DRAWING. 

effect is the formation of increased stresses 
and strains adjacent to the tool that 
strikes the metal at high speed. The 
capacity of most metals to "take" such 
an impact blow is not restricted; therefore 
the blank will be only somewhat less uni 
formly deformed at high rates of working 
than at lower ones. On the contrary, in 
tension-type working, as in drawing rod, 
wire and tubing, and also in rolling sheet 
under tension, the strength of the metal 
limits the maximum force that can be 
transferred from the acting tool to the 
section being worked. Therefore, impact 
consumes a portion of the force that could 
be utilized for an increased reduction to 
be performed in a single operation. 

The methods being used to decrease 
impact effects are mainly of two types. One 
method reduces the speed of the acting tool 
during the starting period, either directly 
by the use of direct current motors or 
indirectly by inserting "soft" members, 
which yield before the maximum force is 
developed. The other method is to design 
the tool in such a manner that the metal it 
self acts as an impact cushion. For example, 



28 



SOME FUNDAMENTALS OF THE FLOW AND RUPTURE OF METALS 



in deep drawing (Fig. 23) sharp-cornered 
tools almost immediately create the maxi 
mum metal resistance and the impact 
effect, while generously rounded off tools 
permit the metal to move before developing 
a high now and impact resistance. As 
previously discussed, the impact effect is 
mainly restricted to the region where tools 
and stock strike each other (Fig, 22). The 
particles in this contact area suffer the 
highest impact acceleration, but this ac 
celeration is rapidly reduced as the distance 
from the contact area increases. The 
stresses set up in the metal and in the tools 
are of two types, the plastic or elastic 
resistance of the metal plus an additional 
stress proportional to the acceleration. 
From this feature of any impact effect, the 
interesting conclusion can be drawn that 
the stock should be placed in the parts of 
the dies that are particularly highly 
stressed, while the less strained part should 
be the impacting tool. 

Regarding the effects produced by the 
heat developed during working, it should 
be recalled that the mechanical work con 
sumed in both cold-working and hot-work 
ing is almost entirely converted into heat. 56 
This heat comes from two sources, a uni 
form or volume part originating from the 
strain within the metal and a surface 
portion resulting from the friction. The 
corresponding increase in temperature is 
determined by the dissipation of the heat 
mainly due to conduction through the tools 
and the metal itself. Therefore, the faster 
the total volume of metal is being worked 
and the shorter the time the tools remain 
in contact with the metal after working, 
the higher will be the increase in tempera 
ture of the metal, other factors being equal. 
This temperature increase can assist the 
working, as in impact extrusion, or it can 
cause difficulties, as in the hot-working of 
a metal with a narrow working temperature 
range. Apparently little has been done as 
yet to apply the laws of heat transfer to 
metal- working problems. 



CONCLUSION 

In concluding, may I be permitted to say 

a few words as to the probable future 

development of the knowledge of the 

plastic deformation and rupture of metals. 

As a science, this branch of metallurgy is 

most akin to the theory of elasticity. 

However, the complex theory of elasticity 

is based on a minimum number of basic 

assumptions, incorporated in Hooke s law, 

and on the knowledge of a few properties 

of the material; i.e., the elastic constants. 

Other factors, such as the deviations from 

Hooke s law, or imperfections of elasticity, 

are of minor importance and do not 

diminish the tremendous practical useful 

ness of the theory of elasticity. On the 

contrary, the plasticity and rupture of 

metals depend, as discussed, upon numer 

ous fundamental factors and also upon 

numerous specific metal properties. There 

fore, extensive research combining care 

fully balanced theoretical and experimental 

work will be necessary to achieve further 

success. Unfortunately, rather sensitive, 

complex and expensive experimental equip 

ment is required for this work, and as a 

consequence the progress in this particular 

field of metallurgy will be rather slow. 

BIBLIOGRAPHY 

i. The literature on this subject is most completely 
assembled in: H. Promm: Grenzen des elastis- 
chen Verhaltens beanspruchter Stoffe (Limits of 
Elasticity in Strained Materials). Handbuch der 



propo 



staende. Leipzig, 1885. Also, Dingler s Polyt. 
Jnl. (1879) 234, 257, 345. 

4. D. J. Me Adam and R. W. Clyne: Proc. Amer. 

Soc. Test. Mat. (1938) 38 (II), 112-134. 

5. T. V. Buckwalter and O. J. Horger: Trans. 

Amer. Soc. Metals (1937) 25, 229-243. 

6. A. Beck: _ Magnesium und seine Legierungen 

(Magnesium and Its Alloys). Berlin, 1939. 

7. T. G. Docherty: Engineering (1935) 139, 211-213, 

285-286. 

8. W. R. Webster, J. L. Christie and R. S. Pratt: 

Trans. A.I.M.E. (1927) Inst. Metals Div.,233- 
252; (1933) 104, 166-169. 

9. H. W. Gillett: Metals and Alloys (1931) 2, 215- 

222. 

10. H. J. Gough and G. A. Hankins: Proc. Inst. 

Automobile Engrs. (1934-35) 29, 543-581. 

11. H. W. Swift: Inst. Automobile Engrs., Sym 

posium on Deep Drawing Research (1940) 
i-73- 

12. M. de St. Venant: Compt. rend. (1870) 70, 

473-480. 

13. M. LeVy: CompL rend. (1870) 70, 1323-1325. 

14. W. Lode: Ztsch. Physik (1926) 36, 913-939. 



GEORGE SACHS 



IS- 

16. 

17, 
18. 
19. 

20. 
21. 
22. 
23. 
24. 
25- 

26. 
27- 

28. 

29, 
30. 
31- 
32. 

33- 

34- 

35- 
36. 



M. Ros and A. Eichinger: Versuche zur Klaerung 
der Frage der Brachgefahr (Experiments on 
the Problem of Failure), Zurich, 1926, 1928. 

R. v. Mises: Nachr. Ges. Wiss. Goettingen, Math. 
Phys. Kl. (1913) 582-597. 

B. P. Haigh: Engineering (1920) 109, 158-160. 
G. Sachs: Ztsch. Ver. D. Ing. (1928) 72, 734-736. 
H. L. Cox and D. G. Sopwith: Proc. Phys. Soc. 

(i937) 49, 134-iSi. 
W. Kuntze and G. Sachs: Ztsch. Ver. D, Ing. 

(1928) 72, 1011-1016. 
G. Bierrett: Mitt. Deutsch. Materialpruef. Anst., 

Sonderheft. No. 15 (1931) 1-39. 
P. W. Bridgman: Trans. A.I.M.E. (1938) 128, 

15-36. 
P. W. Bridgman: Jnl. Applied Physics (1938) 9, 

517-528; Mech. Rug. (1939) 107-1 1 1. 
L. Herrmann and G. Sachs: Metallwirtschaft 

(i934) 13, 745-752. TT 

C. S. Barrett and L. H. Levenson: Trans. 
A.I.M.E. (1939) 135, 327-343. 

G. Sachs: Jnl. Inst. Metals (1939) 64, 261-283. 
T. V. Karman: Ztsch. Ver. D. Ing. (1911) 55, 

1749-1757. 
R. W. Goranson: Bull. Geol. Soc. Amer. (1940) 

51, 1001-1034. 
E. Siebel and A. Maier: Ztsch. Ver. D. Ing. (1933) 

77, 1345-1349. 
E. Siebel and E. Kopf : Ztsch. Metallkunde (1934) 

26, 169-173. 
W. Kuntze: Kohaesionsfestigkeit (Cohesive 

Strength). Berlin, 1932. 

D. J. McAdam and G. W. Clyne: Proc. Amer. 
Soc. Test. Mat. (1938) 38 (II) 112-134. 

M. Gensamer: Metal Progress (1940) 38, 59-64. 
G. Sachs: Iron Age (Aug. i, 1940) 146, 31-34; 

(Aug. 8, 1940) 35-37- 
See C. F. Elam: Distortion of Metal Crystals. 

Oxford, 1935. 
See E. Schmid and W. Boas: Kristallplastizitaet 

(Plasticity of Crystals). Berlin, 1935. 



37. 

38. 
39- 
40. 
41. 
42. 
43- 

44- 
45- 

46. 
47. 
4 8. 
49- 

50. 

5L 

52. 
53- 
54- 
55- 
56. 



Frhr. v. Goeler and G. Sachs: Ztsch. Physik 

(1927) 41, 103-115; Ztsch. Tech. Physik (1927) 

8, 586-594- 
R. Karnop and G. Sachs: Ztsch. Physik (1927) 

41, 116-139. 
G. I. Taylor: Proc. Roy. Soc. (1927) n6-A, 

39-6o. 
B. Chalmers: Proc. Roy. Soc. (1937) i62-A, 

120-127. 
G. I. Taylor: Trans. Faraday Soc. (1928) 4, 



121-125 ; Jnl. Inst. Metals (1938) 62, 307-324. 
(J. S. Barrett: Trans. A.I.M.E. (1940) 137, 

128-145, 
Frhr. v. Goeler and G. Sachs: Ztsch. Physik (1927) 

41, 873-888, 889-906; (1929) 56, 477-484, 

485-494, 495-502. 

J. Weerts: Ztsch. Metallkunde (1933) 25, 101-103. 
0. Dahl and J. Pfaffenberger: Ztsch. Physik 

(1931) 71, 93-1.05. 
N, P. Goss: Trans. Amer. Soc. Metals (1935) 23, 

511-544; (1936) 24, 967-1036. 
R. M, Bozorth: Trans. Amer. Soc. Metals (1935) 

23, 1 107-111 1. 
A. Joffe: The Physics of Crystals. New York, 

1928. 
P. Ludwik: Elemente der technologischen 

Mechanik (Fundamentals of Technological 

Mechanics). Berlin, 1909. 

A. Nadai: Jnl. Applied Mech. (1933) I, 111-129. 
G. Sachs and K. R. Van Horn: Practical Metal 
lurgy, Cleveland, 1940. 

G. Sachs: Spanlose Formung (Mechanical Work 
ing), 38-67. Berlin, 1930. 
A. Portevin and P. G. Bastien: Jnl. Inst. Metals 

(1936) 59, 83-110. 
K. Hanser: Ztsch. Metallkunde (1926) 18, 247- 

255. 
J. Winlock and R. W. E. Leiter: Trans. Amer 

Soc. Metals (193?) 25, 163-205. 
W. S. Farren and G. I. Taylor: Proc. Roy. Soc. 

(1925) IQ7-A, 422-451. 



Internal Friction of Single Crystals of Copper and Zinc 



BY THOMAS A. READ,* JUNIOR MEMBER A.I.M.E. 



(New York Meeting, February 1941) 



THE internal friction of single crystals 
of metals is affected markedly by a variety 
of factors, which, according to the litera 
ture, are without influence on the internal 
friction of polycrystalline metals at strain 
amplitudes less than io~ 5 . For example, the 
internal friction of a suitably oriented zinc 
single crystal is multiplied by a factor of 
100 when the vibration amplitude is 
slightly increased, and the internal friction 
of a copper crystal is markedly increased 
by the application of a stress of 30 Ib. per 
sq. in. for one minute. In this paper it will 
be shown that these, together with the 
other phenomena associated with the 
damping of mechanical vibrations in metal 
single crystals, can be satisfactorily ex 
plained, at least qualitatively, in terms of 
an atomic mechanism that has been pro 
posed for quite another purpose that of 
accounting for slip in metals. The extra 
ordinary agreement between the observed 
behavior of the metal single crystals and 
the behavior that could be predicted on the 
basis of this atomic mechanism leads to 
the belief that the latter is probably essen 
tially correct. If this is so, it follows that 
valuable information about the slip process 
in metals may be gained from measure 
ments of internal friction on single crystals. 
In this paper the results of the only 
study to date on the damping of mechani 
cal vibrations in single crystals of metals 
will be presented. 

Manuscript received at the office of the Institute 
Nov. 29, 1940. Issued as T.P. 1309 in METALS 
TECHNOLOGY, April 1941. 

* Westinghouse Research Fellow, Westinghouse 
Research Laboratories, East Pittsburgh, Pa. 



EXPERIMENTAL PROCEDURE 

The internal friction of a material is a 
measure of the rate at which the energy of 
mechanical vibrations is converted into 
heat. The measure of internal friction used 
in the present paper is the decrement A, 
which is denned as the fraction of the 
energy of vibration lost per half cycle. 
The measurements described here were 
made for longitudinal oscillations at a fre 
quency of 33.5 kc. for the copper crystals, 
and at a frequency of 39 kc. for the zinc 
crystals. 

The method used is that of the composite 
piezoelectric oscillator. The single-crystal 
specimens are in the form of circular cylin 
ders )- in. in diameter and i} to 2% in. 
long. The crystal rods are waxed in a 
V-block and sawed to the proper length 
with a fine jeweler s saw. The ends of the 
specimen are ground flat, and one end is 
cemented with a very thin film of beeswax 
and rosin to the end of an x-cut quartz 
rod which is y in. in diameter and 3 in. 
long. The length of the specimen rod is 
determined by the condition that its 
natural frequency for longitudinal vibra 
tion be the same as that of the quartz rod, 
within 0.5 per cent. The lengths of the 
specimens are thus different for different 
orientations of the cylinder axis with 
respect to the crystallographic axes. When 
the specimen and quartz rods are thus 
matched in frequency the adhesive is at a 
node of stress, and it has been shown experi 
mentally that under these circumstances 



THOMAS A. READ 



the energy dissipation in the adhesive 
material is negligibly small. 

This composite oscillator is mounted in 
a vacuum either horizontally or vertically, 
but in either case it is supported by fine 
wires at nodes of displacement. Two brass 
electrodes are placed on either side of the 
quartz rod at the ends of the electric axis 
(in an x-cut quartz rod the electric axis is 
perpendicular to the axis of the rod). The 
composite oscillator is excited to longitu 
dinal vibration by placing on these elec 
trodes an alternating electromotive force of 
appropriate frequency. A composite piezo 
electric oscillator is shown schematically 
in Fig. r. 

The measurement of the rate at which 
energy is dissipated in the composite oscil 
lator is thus reduced to an electrical meas 
urement, and is carried out by connecting 
the electrodes in one arm of an alternating- 
current bridge. The decrement of the 
specimen rod is calculated from data on the 
variation with frequency of the equivalent 
electrical impedance of the composite 
oscillator in the neighborhood of the 
resonance frequency, and from a knowledge 
of the decrement of the quartz rod. 

A very important feature of this method 
of measurement is that it is possible to 
measure with precision the amplitude of 
stress in the specimen, even though it be 
as small as o.oi Ib. per sq. in. This is done 
by measuring the absolute value of the 
energy dissipated per cycle in the specimen. 
From this and the decrement one can cal 
culate the total energy stored in the me 
chanical vibrations, and thus the amplitude 
of stress in the specimen. The amplitude 
of oscillation at which the measurement 
is made is varied by changing the voltage 
applied to the bridge circuit, and hence 
the voltage on the electrodes of the quartz 
oscillator. 

A complete description of the method of 
the composite piezoelectric oscillator (which 
was developed by Prof. S. L. Quimby, of 
Columbia University, and his students) and 



of the manner in which it has been modified 
for measurements of damping in metal 
single crystals may be found in an earlier 
paper. 1 




SPECIMEN 



QUARTZ ROD 



CfRODES 

FIG. i. COMPOSITE PIEZOELECTRIC OSCIL 
LATOR MADE BY CEMENTING QUARTZ AND 
SPECIMEN RODS END TO END, SUSPENDED BY 
FINE WIRES IN AN EVACUATED CONTAINER. 

SPECIMEN MATERIAL 

The copper crystals on which the meas 
urements were made were grown from Chile 
copper, which contains 99.998 per cent Cu. 
It was obtained in the form of disks y in. 
in diameter and J-{ 6 in. thick and was 
melted in a graphite crucible in a vacuum 
induction furnace and cast into rods. These 
rods were converted into single crystals 
by the method of gradual solidification 
from the melt in a vacuum furnace similar 
to the one described by Nix. 2 

The zinc crystals were grown by the 
method described by Miller 3 from c.p. 
zinc (99.999 per cent Zn) supplied by the 
New Jersey Zinc Co. The principal features 
of this method of production of single 
crystals are as follows: The zinc is melted in 
a Pyrex flask and thoroughly degassed in a 
vacuum. It is then cast in cylindrical molds 
of Pyrex tubing, which have been thor 
oughly cleaned and coated with colloidal 
graphite, to prevent the zinc from sticking 
to the glass. The molds containing the 
polycrystalline zinc are packed in a large 
test tube in Sil-0-Cel and lowered through 
an electric furnace at the rate of y in. 
an hour. 

After growth, the crystals were annealed 
for 2 hr. and cooled slowly to room 
temperature. 



i References are at the end of the paper. 



INTERNAL FRICTION OF SINGLE CRYSTALS OF COPPER AND ZINC 



RESULTS ON COPPER small compression machine was constructed 

If measurements are made on a copper (Fig. 4)- This device was carefully made so 

crystal immediately after it has been that a specimen could be compressed be- 



COPPER SINGLE CRYSTAL 



" 



-NOT ANNEAL D 



STRAIN AMPLITUDE X IQ8 

FIG. 2. INTERNAL FRICTION OF COPPER 
SINGLE CRYSTAL DECREASED BY ANNEALING 
TWO HOURS AT 500 C. 

Upper curve shows dependence of decrement 
on strain amplitude before annealing; lower 
curve, after annealing. 

sawed and its ends ground, it is found to 
have a rather high internal friction, which 
varies markedly with the amplitude of the 
oscillations, as shown by the upper curve 
in Fig. 2. If the crystal is then annealed for 
2 hr. at 500 C., and the measurements re 
peated, the decrement is much lower than 
before annealing, and, moreover, varies 
much more slowly with the amplitude of 
oscillation. This behavior is shown in the 
lower curve in Fig. 2. But even after 
annealing there is still a measurable de 
pendence of the decrement on the ampli 
tude; in Fig. 3 the lower curve of Fig. 2 is 
replotted on a different scale, and carried 
out to larger amplitudes. The decrement 
measurements were carried out to larger 
amplitudes on the annealed crystal than 
on the unannealed crystal, because, for a 
given maximum voltage on the quartz 
oscillator, the amplitude of oscillation is 
inversely proportional to the decrement. 

It is evident from the curves of Fig. 2 
that the internal friction of a copper crystal 
is sensitive to handling. In order to inves 
tigate this effect quantitatively, the follow 
ing experiment was carried out. First, a 




50 



10 20 30 40 

STRAIN AMPLITUDE* 10* 

FIG. 3. LOWER CURVE OF FIG. 2 ON EXPANDED 

SCALE. 




FIG. 4. DEVICE FOR APPLYING COMPRESSIVE 

STRESSES TO SINGLE-CRYSTAL RODS. 

tween two plates accurately parallel to one 
another, and perpendicular to the direction 



THOMAS A. READ 



33 



of motion. The load was measured by the 
amount the spring supporting the lower 
sliding member was compressed, by using 
a sensitive dial gauge. 




STRAIN AMPLITUDE X I0 8 

FIG. 5. INTERNAL FRICTION or COPPER 

CRYSTAL INCREASED BY APPLYING SMALL 
STRESSES TO IT FOR ONE MINUTE 

Lowest curve was obtained before stressing 
crystal; next above it after applying 60 Ib. 
per sq. in.; the third after 120; the fourth after 
150 Ib. per sq. inch. 

Using this device, a single crystal of 
copper was successively subjected to loads 
of 3 Ib., 6 Ib., and 7^ Ib., each for one 
minute. The diameter of the crystal was 
Y in., so these loads produced compressive 
stresses in the crystal of 60, of 120 and of 
150 Ib. per sq. in. Before and after each 
stressing of the crystal its decrement was 
measured as a function of the amplitude 
of oscillation, and the four curves obtained 
are shown in Fig. 5. The lowest curve 
represents the first measurement, and the 
other three follow in order; each applica 
tion of stress produced an increase in the 
internal friction. No change in the length 



of the specimen was observed. With this 
method a change of one part in 30,000 
could have been detected by the resulting 
change in the resonant frequency.* 




STRAIN AMPLITUDE X 10* 

FIG. 6. INTERNAL FRICTION OF COPPER 
CRYSTAL AT 23C. (UPPER CURVE) AND AT OC. 
(LOWER CURVE) . 

Preliminary measurements have also 
been made of the temperature dependence 
of the internal friction of copper crystals 
in the neighborhood of room temperature. 
The data presented in Figs. 2, 3 and 5 were 
obtained at 23C. In Fig. 6, the top curve 
of Fig. 5 is replotted, together with a simi 
lar curve representing measurements made 
on the same specimen at oC. immediately 
after the 23C. run. It is to be noted that 
the decrease in the decrement resulting 



* It is possible that the length of the crystal 
was decreased by more than 0.003 P er cent by 
the stressing but that the Young s modulus 
of the crystal was simultaneously lowered (the 
percentage decrease in Young s modulus being 
twice the percentage decrease in length). This 
possibility could be investigated by stressing 
the crystal in tension. 



34 



INTERNAL FRICTION OF SINGLE CRYSTALS OF COPPER AND ZINC 



from this lowering of the temperature is 
only about 20 per cent. 

In addition to the decrement measure 
ments, data were obtained on another 




STRAIN AMPLITUDE X 10* 

FIG. 7. DECREMENT AND FRACTIONAL 

CHANGE OF RESONANCE FREQUENCY OF LONGI 
TUDINAL VIBRATION OF COPPER CRYSTAL ROD 
PLOTTED AS FUNCTIONS OF STRAIN AMPLITUDE. 

phenomenon, the variation of the resonant 
frequency with the amplitude of oscillation. 
The resonant frequency is defined as the 
frequency of driving voltage that produces 
the maximum amplitude of oscillation. It is 
determined by the length, density, and 
Young s modulus of the specimen, except 
for a small correction depending on the speci 
men diameter. It was found that as the 
amplitude of oscillation was increased the 
resonant frequency of the single-crystal 
rod decreased by a small amount, never 
more than about o.oi per cent. In Fig. 7, 
typical curves showing the dependence of 
decrement and resonant frequency on 
vibration amplitude are reproduced. The 
decrease in the resonant frequency as the 
amplitude is increased is plotted in terms 
of the fractional change in frequency, 

///, from the resonant frequency at a 
very small amplitude. 



RESULTS ON ZINC 

The experimental data on single crys 
tals of zinc have been previously pub 
lished, 1 but they will be reviewed here so 



200 



TIME IN HOURS 

FIG. 8. DECREASE IN INTERNAL FRICTION OF 
ZINC CRYSTAL WITH TIME AFTER MOUNTING. 
All these measurements were made at ampli 
tudes so small that the decrement was inde 
pendent of the amplitude. The gap in the curve 
represents a one-hour anneal at 25oC. 

that they may be compared with the re 
sults on copper. 

As in the copper crystals, the internal 
friction of zinc crystals was high immedi 
ately after the process of sawing the crystal, 
if no annealing process intervened. But an 
additional phenomenon was then ob 
served; while the specimen was mounted 
in the apparatus at room temperature, its 
decrement decreased, at first very rapidly, 
and then more slowly, in the manner shown 
in Fig. 8. After 4 days, corresponding to 
the break in the curve, the composite 
oscillator was removed from the apparatus 
and the zinc crystal annealed for i hr. at 
25oC. It was then replaced in the appara 
tus and the measurements continued. It is 
remarkable that the decrement at the 
beginning of this series of measurements 
was almost a thousand times the final 
value, which was 1.5 X io~ 6 . All of these 
measurements were made at very small 
vibration amplitudes so small that de 
creasing the amplitude by 50 per cent did 
not produce any measurable change in the 
decrement. Because of this dependence of 
the decrement of zinc crystals on the time 



THOMAS A. READ 



35 



after mounting, all the measurements to 
be described were made several days after 
the mounting process, when the decrement 
had reached a steady value. 

It was also found that the properties of 
the crystals are changed by the oscilla 
tions themselves. Fig. 9 shows the way in 
which the internal friction depends on the 
length of time the crystal is oscillated. 
After 10 min., or 25 million cycles, a sen 
sibly constant value is reached, so all read 
ings were taken after this steady state was 
attained. At very small amplitudes this 
effect was not observed. 

Another complication in making meas 
urements of the internal friction of zinc 
crystals is that the values obtained de 
pended on the order in which the measure 
ments are made. This phenomenon is 
illustrated in Fig. 10. Starting the measure 
ments at the smallest vibration amplitude 
and making measurements at successively 
increased amplitudes, the lower branch of 
the curve was obtained. If the measure 
ments at the lower amplitudes were then 
repeated, values of the decrement higher 
than before were obtained, corresponding 
to the upper branch of the curve. 

Finally, it is found that there is a large 
difference in the damping properties of zinc 
crystals whose orientations of the cylinder 
axis with respect to the hexagonal axis are 
different. Fig. n shows the dependence of 
decrement on strain amplitude for several 
crystals of various orientations. The angle & 
is the angle between the cylinder axis and 
the hexagonal axis. 

As with copper, the resonant frequency 
of zinc crystals decreases as the vibration 
amplitude is increased. In Fig. 12 the frac 
tional decrease in resonant frequency is 
plotted as a function of the strain ampli 
tude in the specimen for the same crystals 
whose decrements are plotted in Fig. n. 

This change in the resonant frequency 
can be correlated with the decrement 
changes. If the data of Figs, n and 12 are 
combined by plotting the decrement 



against the change in the resonant fre 
quency (Fig. 13), a straight-line relation 
ship is found. This means, then, that for 
each of these crystals the ratio r = 



I-26X ICT 5 



-20 



10 



20 



30 



TIME IN M .NUTES 

FIG. 9. DECREMENT OP ZINC CRYSTAL ROD 
INCREASES WITH TIME WHILE OSCILLATING. 



100 - X I0~ 5 




2 4 X IO- 7 

STRAIN AMPLITUDE 

FIG. io. DECREMENTS MEASURED WHEN 

STRAIN AMPLITUDE IS INCREASED BY STEPS 
ARE SMALLER THAN THOSE WHEN STRAIN 
AMPLITUDE IS SUBSEQUENTLY DECREASED. 

A -T- j is a constant independent of 

the strain amplitude. The values of r for 
the different crystals are given in Table i. 

TABLE i. Values of Ratio r = A -5- -~ for 
Different Zinc Crystals 



20. O 
61.2 
6I.S 

74-5 
88.0 



2.8 
10.3 
II. 

9.0 

i.i 



An important characteristic of any dis- 
sipative process, a knowledge of which is 
essential to an investigation of the origin of 
the dissipation, is the dependence on fre 
quency. Zinc single crystals are so sensitive 
to handling and other influences that are 
difficult to control that it is necessary that 
the measurements at the different fre 
quencies be made on the same specimen, 



INTERNAL FRICTION OF SINGLE CRYSTALS OF COPPER AND ZINC 



and with all other factors kept constant. 
This can be done by making measurements 
on the same composite oscillator, but at its 
second harmonic. A typical result is given 



the hexagonal axis, and the smallest ten 
sion will be required to produce slip in a 
single-crystal rod if the basal plane makes 
an angle of 45 with the axis of the rod 




50 100 

STRAIN AMPLITUDE X I0 7 

FIG. ii. VARIATION or DECREMENT WITH STRAIN AMPLITUDE IN ZINC CRYSTALS WHOSE CYLINDER 

AXES MAKE VARIOUS ANGLES Q WITH HEXAGONAL AXIS. 



in Fig. 14, which shows that the decrement 
for a given strain amplitude is approxi 
mately inversely proportional to the 
frequency. 

DISCUSSION OF RESULTS 

There are two striking differences be 
tween the behavior of copper crystals and 
that of zinc crystals. One is that there is no 
analog of Figs. 8 and 9 for copper, and the 
other is that only in the case of zinc is 
there a marked dependence of the decre 
ment on the orientation of the crystal. 
This second difference is very suggestive, 
if we keep in mind that the slip properties 
of the two metals differ in just the same 
way. Zinc, with its hexagonal crystal 
structure, slips at room temperature 
only on the basal plane perpendicular to 



The crystal structure of copper, on the 
other hand, is face-centered cubic, and it 
will slip on all the (in) planes. For any 
orientation of the rod one of these planes 
will lie at or near the 45 position. 

Further evidence that the damping of 
mechanical vibrations in zinc crystals re 
sults from their plasticity is afforded by the 
observed dependence on frequency. An 
inelastic deformation proportional to the 
length of time the stress is applied gives 
rise to a decrement inversely proportional 
to the frequency, like that observed for 
zinc crystals. 

The conclusion is that the phenomena 
of slip and the dissipation of vibrational 
energy are related. Before discussion of 
this relationship, let us digress to consider 
the dislocation theory of slip in metals. 



THOMAS A. READ 



37 



DISLOCATION THEORY OF SLIP* 
Metals are deformed plastically by the 
processes of slip and of twinning. By 
"slip" is meant the relative displacement 



posed that the elementary unit of slip is 
the relative displacement of two planes in 
the crystal, each plane moving as a unit. 
An estimate of the shearing stress neces- 




150 



200 



50 100 

STRAIN AMPLITUDE X I0 7 

FIG. 12. FRACTIONAL VARIATION OF RESONANCE FREQUENCY WITH STRAIN AMPLITUDE IN ZINC 
CRYSTALS WHOSE CYLINDER AXES MAKE VARIOUS ANGLES WITH HEXAGONAL AXIS. 



of two planes in a crystal lattice by a dis 
tance equal to some integral multiple of 
the distance between two equilibrium 
positions for atoms in that plane. After 
slip has taken place, the atoms are in new 
equilibrium positions, and the ideal lattice 
structure is preserved. Slip may occur 
simultaneously on a great many parallel 
crystallographic slip planes. The slip plane 
in a metal usually is the plane that is most 
thickly populated with atoms, and the slip 
direction is parallel to the line in the plane 
that has the greatest atomic density. 
Visual evidence for the existence of slip 
planes is given by the formation of glide 
ellipses on the surface of a plastically de 
formed single-crystal rod ellipses formed 
by the intersection of the cylindrical sur 
face and the plane of slip. 

Since the slip process has such a definite 
crystallographic nature, it might be sup- 

* A good review of this subject appears in chapter 
V of "Elasticity, Plasticity, and Structure of Matter" 
by R. Houwink. Cambridge Univ. Press, 1937. 



sary to produce such a motion can be made 
easily. When the plane is just halfway be 
tween two equilibrium positions the shear 
is approximately J4? and the shearing 
stress is roughly Y^G, where G is the shear 
modulus. An average value of G for metals 
is 5 million Ib. per sq. in., so the calculated 
stress necessary to produce slip is about 
i million Ib. per sq. in. In order to account 
for the discrepancy between this figure and 
the observed values, which are in the 
neighborhood of 100 Ib. per sq. in., slip 
mechanisms have been proposed, according 
to which slip takes place over only a small 
part of the slip plane at once. One way in 
which this can happen is through the 
motion of dislocations through the crystal 
lattice. A dislocation, which is a particular 
type of imperfection in the regularity of 
the crystal lattice, is illustrated in Fig. 15. 
When such a dislocated region moves from 
one side of the crystal to the other, the 
two atomic planes undergo a relative dis- 



INTERNAL FRICTION OF SINGLE CRYSTALS OF COPPER AND ZINC 



placement of one lattice distance. It has 
been illustrated by Taylor 4 that the atoms 
in the dislocation are bound relatively 



mobility of the dislocations will be en 
hanced, and some of them will migrate out 
of the crystal, or will coalesce with dis- 



m cne dislocation are uuuim ieia,uvciy ui LJ.J.C ^i_yaca.j. ? ^j. WJ..LJ. \^\JU.L^O\^^ YV.II.JU ujg- 
loosely to their positions of equilibrium; locations of the opposite sign. After the 



200 



UJ 

lOO 




X I0 5 

FIG. 13. COMBINATION OF DATA PLOTTED IN FIGS, n AND 12. 
Shows that ratio of decrement to fractional change in frequency is constant. 



for this reason the dislocation can be 
moved through the crystal by a very small 
stress <TI. In general, the stress cr 2 required 
to produce the dislocation will be larger 
than <TI, and will be the stress it is necessary 
to apply to the crystal in order to make it 
slip. Let us now apply these ideas to the 
interpretation of the results on the damping 
of mechanical vibrations in metal crystals. 

INTERNAL FRICTION OF METAL CRYSTALS 

Since no actual crystal is perfect, every 
crystal will contain a certain number of 
dislocations, depending on how much 
plastic deformation it has undergone. 
Under the influence of an oscillating stress, 
these dislocations will move back and 
forth in the crystal and thereby dissipate 
the energy stored in the oscillations. In 
moving the dislocations, the external stress 
will be aided by the random fluctuations 
in stress in the crystal that arise from the 
thermal motions of the atoms. If the crystal 
is raised to an elevated temperature, the 



crystal has been brought back to room 
temperature, the number of dislocations in 
the crystal will be smaller and the internal 
friction lower. This prediction is verified by 
Fig. 2 for copper and by Fig. 8 for zinc. For 
zinc, we find that the dislocations migrate 
out of the crystal at a finite rate at room 
temperature. 

As the converse of annealing, deforming 
the crystal plastically increases the den 
sity of dislocations in the crystal; this 
should increase the internal friction. Fig. 5 
gives the result of such an experiment on 
copper, and shows that the internal friction 
is raised by slight amounts of cold-work. 
With zinc, it is not necessary to apply such 
large shearing stresses to the slip plane; 
the internal friction is increased measurable 
amounts by the alternating stress asso 
ciated with the vibrations when this stress 
is as low as i Ib. per sq. in. (Figs. 10 and 1 1) . 

Any theory of the dissipation process in 
these metal crystals that is worth while 
must offer an explanation of the depend- 



THOMAS A, READ 



39 



ence of the decrement on the amplitude of 
vibration. This is done by the mechanism 
of the motion of dislocations in the follow 
ing way: The stress 0% which is required 



decrement-amplitude curve does become 

flatter at strain amplitudes above 4 X icr 6 . 

This discussion applies to both copper 

and zinc, but for zinc there is an additional 




5 10 15 

STRAIN AMPLITUDE X I0 7 

FIG. 14. INTERNAL FRICTION OF ZINC CRYSTALS is VERY NEARLY INVERSELY PROPORTIONAL TO 

FREQUENCY AT ALL STRAIN AMPLITUDES. 

Vibration frequency for data of upper curve is 39 kc.; of lower curve, 78 kc. 



to move a dislocation through the lattice, 
is in general larger than the oscillating 
stress associated with the vibrations. Mo 
tion of the dislocations is made possible by 
the thermal vibrations of the atoms in the 
lattice, which may be regarded as pro 
ducing fluctuations in the stresses that 
act on the dislocations. As the amplitude 
of vibration is increased, the additional 
stress that must be supplied by the thermal 
motions in order to make the total stress 
acting on the dislocations equal to <TI be 
comes smaller, so the rate of motion of the 
dislocations is greatly increased. If this 
explanation is correct, it would be ex 
pected that the dependence of decrement 
on stress amplitude would not be so great 
if the vibrational stress exceeds <TI. This 
prediction may be compared with the 
data shown in Fig. 3, which shows the 
result of measurements at relatively large 
amplitudes and demonstrates that the 



cause of the amplitude dependence. As the 
amplitude is increased, more dislocations 
are produced in the zinc crystals, which 




POSITIVE DISLOCATION 




d 6 f 

NEGATIVE DISLOCATION 

FIG. 15. FORMATION AND PROPAGATION OF A 
DISLOCATION (ATTER TAYLOR). 

gives rise to a rapid increase in their damp 
ing even up to the largest amplitudes of 
oscillation. 



INTERNAL FRICTION OF SINGLE CRYSTALS OF COPPER AND ZINC 



Another feature of the measurements, 
which can be understood in terms of the 
dislocation picture, is the change of 
resonant frequency produced by increasing 




FIG. 16. SUGGESTED FORM OF STRESS-STRAIN 
DIAGRAM FOR SINGLE METAL CRYSTALS. 

the amplitude of vibration. It is important 
to realize that this effect could not be 
produced by an inelastic strain rate 
uniquely determined by the stress; i.e., 
independent of the time. It can be pro 
duced only by some process that gives rise 
to an asymmetrical stress-strain loop. Such 
a process is provided by the motion of dis 
locations, at least under certain circum 
stances. If it is assumed that during the 
quarter cycle in .which the stress increases 
from zero to its maximum value a large 
proportion, say 50 per cent, of the available 
dislocations are moved until they are 
stopped by some obstruction such as an 
impurity atom, in the next quarter cycle 
fewer dislocations are moved, as most of 
them will have been "used up." This situa 
tion has been idealized in Fig. 16. The line 
oe represents purely elastic behavior and 
gives the actual stress-strain relationship 
at very small stresses. At larger stresses an 
inelastic strain is produced as the result 
of the motion of dislocations represented 
by the strain db. Assume now that as the 
stress decreases no further motion of dis 
locations takes place, and part be of the 



loop is simply an elastic contraction. It 
must be remembered that Fig. 16 is not 
drawn to scale. The actual plastic strain, 
shown by the intercept ac, is not larger 
than one-thousandth the strain amplitude, 
shown by the abscissa U/i. The stress- 
strain loop is extremely narrow, and the 
slope of the line ob may be taken as the 
effective Young s modulus. It can be 
shown that for the loop illustrated in Fig. 
1 6 the ratio r of decrement to fractional 
change in frequency lies between 4 and 8, 
depending on the shape of the curve ab and 
is independent of the amplitude of the 
stress. This result is in general agreement 
with the data of Table i. 

SUMMARY 

1. Measurements of the internal fric 
tion of single-crystal rods of copper and 
zinc have been made for longitudinal oscil 
lations at frequencies of 33.5 and 39.0 kc., 
respectively. The measurements on zinc 
were also made at 78.0 kilocycles. 

2. By careful annealing, the decre 
ments of the single-crystal rods may be 
made as low as i X io~ 5 . 

3. The internal friction of single crys 
tals of copper and zinc is increased by even 
very small amounts of cold-work. 

4. The decrements of all the crystals 
vary markedly with the amplitude in the 
range of "small strains." 

5. The damping properties of copper 
crystals do not change with time at room 
temperature, and are not affected by vibra 
tions of the amplitude employed. 

6. The zinc crystals exhibit the phe 
nomenon of recovery at room temperature. 
Their internal friction is increased by all 
but the smallest oscillating stresses. 

7. The internal friction of zinc crystals 
depends in such a manner on the angle 
between the rod axis and the hexagonal 
axis as to indicate that the rate at which 
the energy of vibration is dissipated is 
determined primarily by the resolved shear 
stress on the slip plane. 



DISCUSSION 



8. The resonant frequencies of the 
crystal rods change as the vibration 
amplitude is increased in a way that can be 
correlated with the decrement changes. 

9. The internal friction of zinc single 
crystals is approximately inversely pro 
portional to the frequency. 

10. The behavior of these crystals can 
be explained in terms of the dislocation 
theory of slip. 

ACKNOWLEDGMENT 

The writer wishes to thank the New 
Jersey Zinc Co. for supplying the pure zinc 
used in these measurements. 

REFERENCES 

1. T. A. Read: Phys. Rev. (1940) 58, 371. 

2. F. C. Nix: Rev. Sci. Instr. (1938) 9, 426. 

3. R. F. Miller: Trans. A.I.M.E. (1936) 122, 176. 

4. G. I. Taylor: Proc. Roy. Soc. (1934) I4S-A, 362. 

DISCUSSION 

(H. W. Russell presiding) 
J. T. NORTON,* Cambridge, Mass. Dr. 
Read has done a splendid piece of experimental 
work in examining these single crystals. It is 
really the first complete contribution on the 
internal friction in single crystals, and he has 
given a logical and practical explanation of one 
of the possible causes for internal dissipation of 
energy. The dislocations of Taylor s are well 
suited to explain some of the aspects of internal 
energy dissipation. We have all speculated 
about it more or less, but here seems to be a 
more or less quantitative confirmation of these 
ideas. It is quite possible, however, that this is 
not the whole story. 

Clarence Zener has developed what seems to 
be an adequate explanation of the internal 
energy dissipation on the basis of thermal cur 
rents set up as the crystals in polycrystalline 
materials are deformed elastically. Strangely 
enough, the magnitude of the energy dissipa 
tion that he found in Ms experiments is almost 
identical with that found by Dr. Read. 

It seems to me that in the actual case of the 
energy dissipation in a polycrystalline material 
both of these phenomena play a part, and it is 
difficult to see at the moment how much of a 



contribution comes from the dislocations and 
how much comes from something that is in the 
nature of a nonreversible flow of heat from one 
part of the specimen to another. I sometimes 
wonder whether these are not aspects of the 
same thing. Another possibility that has been 
considered but that has very little experimental 
basis in terms of metals is the possibility of 
some phenomenon related to the primary creep, 
the reversible part of the creep phenomenon. 
It would be possible to account for energy 
dissipation in carrying a crystal through a stress 
cycle on some such basis. For some of the 
plastic materials, where the primary creep can 
be measured accurately and at the same time 
a damping test can be made, these two seem. 
to check out very nicely; so there is another 
aspect of the problem. With more experimental 
work of the kind done by Dr. Read, a much 
more consistent picture of the whole phe 
nomenon should be obtained. 

I think there is some confusion as to the 
units used to describe internal dissipation of 
energy, damping capacity, specific damping, 
and logarithmic decrement. All of those terms 
are in common use and it would be very helpful 
if Dr. Read would explain at least the various 
orders of magnitude involved, and whether 
they are all more or less the same order of 
magnitude or quite different. 

I also wish he could tell us a little bit more 
about the actual electrical method of measuring 
the internal-energy dissipation. 

One other question I have is a matter of 
technique. When Dr. Read makes a composite 
oscillator of the type he describes, how impor 
tant is it to have the quartz and the specimen 
of the same fundamental frequency, and how 
important is it to have the areas of the two 
crystals alike? He shows one square and one 
round. Is it important to have the areas of the 
cross section exactly alike or does not that 
make any difference? 

W. KAUZMANN,* East Pittsburgh, Pa. No 
description is given in the paper of the very 
striking demonstration given by the author 
during the oral presentation at the February 
meeting, of the effect of small amounts of cold- 
work on the damping of vibrations in an 
annealed copper bar. Inclusion of such a 



* Associate Professor, Physics of Metals, Massa 
chusetts Institute of Technology. 



* Westinghouse Research Fellow, Research Labora 
tory, Westinghouse Elec. and Mfg. Co. 



INTERNAL FRICTION OF SINGLE CRYSTALS OF COPPER AND ZINC 



description in the author s closure would be of 
interest. 

The low temperature coefficient of the 
internal friction indicates that the activation 
energy for the motion of dislocations must be 
small (of the order of the average thermal 
energy, or 600 cal per mol). If the author s 
interpretation of the "shelf" in Fig. 3 is 
correct, it should be possible to estimate the 
size of a dislocation from the activation energy 
and the magnitude of the stress at which 
the leveling off of the decrement vs. amplitude 
curve takes place. Has the author carried out 
any such analysis? 

C. S. SMITH,* Waterbury, Conn. The 
author is to be congratulated on a most inter 
esting paper. For the benefit of engineers, it 
might be pointed out that internal friction is 
nothing but an easily obtained and extremely 
sensitive index of the area of a closed stress- 
strain loop. Many of the effects shown by 
internal-friction measurements can also be seen 
on stress-strain curves made with the ordinary 
instruments of the testing engineer if sufficient 
care is taken. For example, in a recent paper 5 
describing tests on 70-30 brass, it has been 
shown that slight overstrain as little as 
0.005 P er ceirt permanent set will produce 
measurable curvature of the stress-strain line 
at stresses far below those at which annealed 
material ceases to obey Hooke s law. Most 
copper alloys behave in a like manner, although 
some precipitation-hardened alloys (beryllium 
copper) seem to be immune from damage by 
plastic deformation. 

C. S. BARRETT,! Pittsburgh, Pa. Bruce 
Chalmers has published measurements on creep 
that show a phenomenon he calls "micro- 
creep," which is a reversible creep at low stress 
levels. He attributes this to the movement of 
dislocations. The phenomenon and its explana 
tion appear to be closely related to the inter 
esting results of this paper, and I should like 
to hear Dr. Read s comments on it. 

D. L. MARTINA New Haven, Conn. The 
author has clearly demonstrated the application 
of internal friction as a metallurgical research 

* Research Metallurgist, American Brass Co. 

5 C. S. Smith: Proportional Limit Tests on Copper 
Alloys. Proc. Amer. Soc. Test. Mat. (1940) 40. 
86^-884. 

T Metals Research Laboratory, Carnegie Institute 
of Technology. 

t Hammond Laboratory, Yale University. 



tool by his correlation of damping and plasticity 
of single crystals. The relation of internal 
friction to the dislocation slip mechanism 
can be extended to explain several phenomena 
that have been observed for polycrystalline 
metals and alloys. By this theory of internal 
friction solid solution alloys would be expected 
to possess a lower internal friction than the 
solvent metals, since the presence of foreign 
atoms in the parent lattice makes the motion 
of dislocations more difficult, and, as a result, 
should decrease the damping capacity. Accord 
ing to Forster and Koster, 6 alloys do have a 
lower damping capacity than their components, 
substantiating this reasoning. 

The author states that the copper crystals 
do not exhibit the recovery phenomenon at 
room temperature. Koster and Rosenthal 7 
found that the damping of drawn polycrystal 
line brass decreases with time at room tempera 
ture in a manner similar to that shown for 
zinc in Fig. 8. Table 2 gives the results for a 
10 per cent and a 60 per cent drawn brass rod, 
showing that the heavy reduced material shows 
the more pronounced drop. This being the case, 
it would be of interest to know whether an 

TABLE 2. Influence of Time at Room Tem 
perature on Damping of Drawn Brass 



Reduction 
Brass Rods, 
Per Cent 


Time, Hours 





10 


20 


100 


Damping io" 4 


10 

60 


S-i 
15-5 


2.5 

5-5 


2.4 
5-0 


2.0 

3.0 



After Koster and Rosenthal. 

elongation of the copper crystal would not 
produce an effect similar to that of the sawed 
zinc crystal. 

An analogous example of recovery at room 
temperature is shown for pure aluminum 
(Fig. 17). Forster and Breitfeld, 8 in deter 
mining this time-damping curve, were able to 
obtain absolute values in 3 or 4 sec. with an 
accuracy of about 2 per cent. It was thus 
possible to follow the damping of the pure 



6 Forster and Koster: Ztsch. Metallkunde (1937) 
29, 116. 

7 Koster and Rosenthal: Ibid. (1938) 30, 345. 

8 Forster and Breitfeld: Ibid. (1938) 30, 343. 



DISCUSSION 



43 



aluminum rod with an initial 6 kg. per sq. mm. 
load during an increase in the load of 0.25 kg. 
per sq. mm. (365 Ib. per sq. in.). This increase 



value as an upper limit to the thermal current 
contribution to the damping. 

The two commonest measures of internal 



was applied over a time of 45 sec. and resulted friction or damping capacity (these two terms 




7 997J 



712.5 



Minutes 



FIG. 17. DAMPING OF ALUMINUM DURING INCREASING LOAD, AND RECOVERY. (AFTER FORSTER AND 

BREITFELD.) 



in a tenfold increase in damping. After the 
loading, the damping immediately decreased 
exponentially to a value somewhat higher than 
the initial one. 

T. A. READ (author s reply). In many cases 
it is possible to discriminate between the con 
tributions to the damping of a metal that arise 
from plasticity and from the flow of thermal 
currents. The latter contribution has been 
thoroughly investigated by Zener, 9 as referred 
to by Professor Norton. The separation is 
based on the fact that the thermal current 
damping is independent of the amplitude of 
vibration. Consequently, if measurements are 
made of the damping over a range of ampli 
tudes, we may take the smallest observed 



9 C. Zener and R. H. Randall: Trans. A.I.M.E. 
(.1940) 137, 41- 



are synonymous) are the specific damping D 
and the logarithmic decrement A. They are 
related by 

D = ^ = 2 A 

where AW is the energy dissipated per cycle 
in the specimen and W is the maximum poten 
tial energy stored in the specimen during the 
cycle of vibration. 

A complete description of the method of 
measurement will be found in an earlier paper 
by the author (ref. i). It is not necessary that 
the area of the quartz and specimen be the 
same. On the other hand, the frequencies of the 
specimen and the quartz must be the same to 
about 0.5 per cent, as mentioned in the paper, 
for otherwise there is a contribution to the 
damping from the adhesive film between the 
specimen and quartz. 



44 



INTERNAL FRICTION OF SINGLE CRYSTALS OF COPPER AND ZINC 



As mentioned by Dr. Barrett, Chalmers has 
suggested the movement of dislocations as a 
mechanism to account for the microcreep he 
observed in single crystals of tin. There are, 




AMOUNT OF COLD WORK 

FIG. 1 8. SCHEMATIC ILLUSTRATION or DE 
PENDENCE OF INTERNAL PRICTION OP COPPER 
ROD ON AMOUNT OP COLD-WORK IT HAS 
RECEIVED. 

The maximum of the curve corresponds 
roughly to dropping the annealed rod on a wood 
surface from a height of several inches. 

however, two striking differences between 
microcreep and the damping of metal single 
crystals. First, the initial rate of microcreep is 
proportional to the stress up to 100 grams per 
sq. mm., whereas over the same stress range 
the damping increases rapidly with the stress. 
Note that an inelastic strain rate proportional 
to the stress should give rise to a decrement 
independent of the amplitude. Secondly, 



Chalmers states that the amount and rate of 
microcreep is insensitive to slight amounts of 
cold-work. Internal friction, on the other hand, 
is affected markedly even by handling the 
specimen. 

The effect of cold-work on the internal fric 
tion of copper may be demonstrated easily 
on a polycrystalline specimen of large grain 
size, such as a cast rod. If such a specimen is 
annealed 30 min. at 25oC. and slowly cooled 
to room temperature, it will have very low 
damping, as may be shown by suspending it 
by a thread and tapping it lightly with a mallet. 
Dropping the specimen on a wooden surface 
from a height of 3 in. suffices to remove the 
effect of this annealing completely; it is then 
"dead soft" and cannot be made to ring 
at all. Severe cold-work, however, lowers the 
damping to a value comparable to that after 
the anneal. This dependence of the internal 
friction on the amount of cold-work is repre 
sented schematically in Fig. 18. 

An estimate of the size of a dislocation has 
been made from internal-friction data, as 
suggested by Dr. Kauzmann. The value of 
qAl, the measure of the size of a dislocation 
used by Dr. Kauzmann 10 obtained in this 
manner is 1.6 X io 4 A. 3 This agrees in order of 
magnitude with the values he has obtained 
from an analysis of creep data. 



10 This volum page 57. 



Time and Temperature Effects in the Deformation of Brass 

Crystals 

BY H. L. BCTRGHOFF* AND C. H. MATHEWSON,t MEMBERS AJ.M.E. 
(New York Meeting, February 1941) 



THE study of the creep of metals under 
conditions of prolonged loading has re 
ceived the attention of many investigators 
for several years and almost innumerable 
papers have been published on the various 
aspects of the subject. As it appeared that 
much could still be learned from a study of 
single crystals, the present work was car 
ried out to study the effects of temperature 
and conditions of loading upon the prop 
erties and deformation in tension of single 
crystals of alpha brass containing about 70 
per cent copper and 30 per cent zinc. In 
particular, it was hoped to obtain informa 
tion from tests below and above the recrys- 
tallization temperature of the material, so 
that properties such as elastic limit, yield 
point or critical stress, and manner of 
deformation might be compared for tensile 
tests at an ordinary loading rate and for 
prolonged loading. 

Accordingly, short-time tensile tests and 
creep tests were made at room temperature, 
300, 500 and 7ooF. The two latter tem 
peratures are above the lowest possible 
recrystallization temperature of polycrys- 
talline 70-30 brass. 

PREVIOUS WORK 

Clarke and White 1 have made creep tests 
of some copper-zinc alloys, including 70-30 

From a part of a dissertation presented by H. L. 
Burghoff to the Faculty of the School of Engineering 
of Yale University in partial fulfillment of the require- 



Manuscript received at the office of the Institute 
Nov. 27, 1940. Issued as T.P. 1288 in METALS 
TECHNOLOGY, February 1941. 

* Research Metallurgist, Chase Brass and Copper 
Co., Inc., Waterbury, Conn. 

f Professor of Metallurgy, Yale University, New 
Haven, Conn. 

1 References are at the end of the paper. 



brass, and copper-zinc-tin alloys at tem 
peratures both below and above the recrys 
tallization temperature of the materials 
considered. Hanson and Wheeler 2 have 
studied the flow and fracture of single 
crystals of aluminum for short-time and 
prolonged tensile loading over a range of 
temperature extending from room tempera 
ture to 4ooF., their experiments being 
made without consideration of the orienta 
tions of the test specimens. Extensive creep 
and fracture tests on single crystals of lead 
at room temperature have been made by 
Betty 3 and by Baker, Betty and Moore. 4 
Results of the last-named investigation 
were published while the present work was 
in progress. Miller 5 examined the elastic 
limit of zinc single crystals at and above 
room temperature by means of creep tests, 
while Miller and Milligan 6 reported the 
results of tests designed to show the influ 
ence of temperature on elastic limit of 
single crystals of aluminum, silver and zinc. 
The behavior of iron crystals under static 
tensile loading at room temperature was 
reported by Gensamer and Mehl 7 while the 
present investigation was in progress. Their 
specimens exhibited a well-defined yield 
point, below which no appreciable creep 
could be detected. The manner of yielding 
proved interesting, as the shape of the 
creep-time curve for the iron crystals dif 
fered from the common type of creep 
curves. Elongation did not begin imme 
diately, or at least was initially very slow. 
The rate of elongation then increased con 
tinuously and finally decreased to zero. 
This S-shaped type of elongation-time 



45 



TIME AND TEMPERATURE IN DEFORMATION OF BRASS CRYSTALS 



curve was reproduced at progressively 
higher loads and was confirmed for several 
specimens. 

PREPARATION OF MATERIALS 

The crystals were made by the Bridgman 
method in the form of rods of J^-in. diam 
eter and about 9 or 10 in. long. Material for 
the crystals was high-grade 70-30 brass 
containing the following maximum amounts 
of impurities: 0.006 per cent Pb, 0.008 per 
cent Fe, and 0.002 per cent Ni. The finished 
crystals contained about 71 per cent Cu. 

In preparation for testing, the specimens 
were carefully machined to a diameter of 
about 0450 in. to permit a gauge length of 
6 in., annealed for 16 hr. at i35oF. to 
reduce coring, and then cleaned or etched in 
nitric acid solution. 

Orientations of specimens before and 
after testing were determined by Gren- 
inger s back-reflection Laue method. 8 The 
octahedral plane of maximum shear stress, 
the active slip plane, was determined by 
the Taylor and Elam method 9 and the 
tangential component of stress along this 
plane was calculated from Elam s formula: 10 

p 

S = sin X cos X 
A 

where S = resolved shear stress, 
P applied load, 

A = area of cross section of speci 
men, 

X = angle between {in} and speci 
men axis, 

X = angle between [101] and speci 
men axis. 

The theoretical final position of the speci 
men axis (final orientation) was also deter 
mined for comparison with the position 
actually observed, the calculation being 
made from Schmid and Boas formula: 11 

V \r ^0 

sin A i = sin A o T- 
h 

where Xi = final angle between slip plane 
(in) and specimen axis, 



XQ = original angle between slip 

plane { 1 1 1 } and specimen axis, 

/o original length of specimen = 

i 
1 1 final length of specimen = 

i + total elongation. 
Knowing the angle between the slip plane 
and the final position of the specimen axis, 
and the course of movement of the speci 
men axis on the stereographic projection 
according to the Taylor and Elam theory, 
the theoretical position was readily plotted, 

EXPERIMENTAL WORK 

Most of the tests concerned with the 
effect of temperature upon properties of the 
crystals were performed with the equip 
ment devised by Miller 5 - 6 for his work on 
crystals of zinc, aluminum, and silver. In 
this equipment the specimen was suspended 
in a heavy steel frame and connected at the 
bottom to a balanced 6 to i lever. Testing 
temperatures above room temperature 
were attained by means of an electrically 
heated vertical tube furnace which sur 
rounded the test specimen. The tem 
perature was controlled by means of an 
automatic potentiometer controller. 

Loading of the specimen was accom 
plished by allowing sand to run slowly from 
a reservoir into a bucket suspended from 
the longer arm of the 6 to i lever. The 
weight of sand applied was readily meas 
ured by means of a Toledo scale, which 
supported the sand reservoir and was 
graduated to read to o.i Ib. The rate of 
loading was approximately constant at 
about 7 Ib. per minute, and readings of 
load, extension, and specimen temperature 
were taken for increments of 12 or 24 Ib. as 
desired. At critical periods in some of the 
tensile tests, readings were taken at more 
frequent intervals. 

Extension of the specimen was measured 
by means of a microscope extensometer 
with which it was possible to make readings 
to an accuracy of 6.6 X io~~ 6 in. per inch. 
Errors of reading due to thermal expansion 



H. L. BURGHOFF AND C. H. MATHEWSON 



47 



and contraction of the sleeve supporting 
the microscopes of the extensometer were 
eliminated by checking readings for the 
distance between the two by a standard 
quartz rod. Inasmuch as specimen tem 
perature could not be maintained at 



iF. was definitely significant, for it meant 
a difference of approximately two divisions 
of the filar micrometer eyepiece. Tests 
made at room temperature with this equip 
ment were without benefit of temperature 
control, but specimen temperature was 



TABLE i. Data and Test Results for Crystals Tested 













Critical 














Young s 


Max. 






Specimen 

No. 


Xo 
(PA {in}), 
Deg. 


Xo 
(PA [ioi]), 
Deg. 


sin X cos \o 


Modulus of 
Elasticity, 
Grams per 
Sq. Mm. 


Resolved 
Shear 
Stress, 
Grams per 


Total 
Extension, 
Per Cent 


Resolved Creep 
Stress, Grams 
per Sq. Mm. 












Sq. Cm. 






Room Temperature (70-8oF.) 



35-1 


43*2 


49 


0.451 




1,465 


12.5 




14-2 


33 


43 iy 


0-397 




1,440 


7 




31-1 


S3 


58*2 


0.416 




1,400 


7 




I2-I a 


28 


30 


0.406 




1,370 


38 




29-I a 
8-4 


47 
33*2 


48 
32* 


0.489 
0.452 


14.1 X io 6 
6.8 X io 


1,395 
1,300 


13 
0-5 




12-3 
32-2 
28-1 


39 ix 

4SH 

41 * 


39* 
46 
43 


0.485 
0.496 
0.484 


8.5 X io 




12.7 
5-o 


7505900; 1000; 1 100 
1200; 1300 
IIOO 



300F. 



8-3 


2 3 * 2 


35 


0.327 


12.7 X io 6 


1.035 


o. 29 




1 1-3 


30*2 


31 


0-434 


7.7 X IO 6 


1,220 


3.0 




30-2 


4* * 


43 


0.484 






10 


800; 1000; noo; 
















1 200 


30-3 


52 


S6H 


0-435 


12.7 X io 6 


1,100 


7.3 


1000; noo 



500F. 



14-1 


39* 


46 


0.441 


18.5 X io 6 


1,150 


9 




1 1-4 


42 H 


42 H 


0.497 


5-i X io 6 


1,140 


6 




30-1 


35H 


42 


0.430 


14.9 X io 6 




4 


300; 450; 600; 800; 
















1000; noo 


29-2 


39* 


43 H 


0.460 


3.1 X io 




II 


1000; noo 



700F. 



i 
12-4 


37 
46 


38 

42 


0.474 
0.534 


9.4 X io 6 
5.6 X io 


1,290 
1,420 


4 
13 




7-4 


4i*| 


47 


0.451 






0.22 


200; 300; 450 


31-3 




4iH 


0.466 


4.0 X 106 




0.72 


600 



1 See reference 13. 



absolutely constant value during any test, 
it was essential to correct observed readings 
of extension for variations in temperature. 
Specimen temperature therefore was deter 
mined at each step in a test. Instantaneous 
specimen temperatures were read to iF. 
and corrections to be applied to the ob 
served extension were calculated using 
instantaneous coefficients of expansion 
derived from National Bureau of Standards 
data. 12 A change in temperature of only 



recorded with each reading of extension as 
usual. For all tests above room temperature 
auxiliary thermocouples were wired se 
curely to the specimen near each reference 
mark for the purpose of adjusting furnace 
heating to obtain uniform temperature 
distribution in the specimen. 

A number of tensile tests were made at 
room temperature with a hydraulic tensile 
testing machine. A stress-strain recorder 
drew the load-elongation curve up to 0.5 



TIME AND TEMPERATURE IN DEFORMATION OF BRASS CRYSTALS 



per cent elongation in the 2-in. gauge 
length required with this apparatus. 

Tests with the weight and lever loading 
device were made at room temperature 



az 

Axial Strain -percenr 

FIG. i. RESOLVED STRESS-STRAIN CURVE 

FOR SPECIMEN 8-4 IN TENSION AT ROOM 
TEMPERATURE. 

(70 to 8oF.), 300, 500 and 7ooF., and 
included slow tensile tests, in which the 
specimens were extended only until appre 
ciable plastic deformation occurred, and 
prolonged tests at constant load for each 
temperature. 

DISCUSSION OF RESULTS 
Tests at Room Temperature 

Data concerning the specimens tested at 
room temperature, as well as some test 
results, are shown in Table i. 

The several specimens subjected to short- 
time tests in the tensile testing machine 
showed a definite yield point; i.e., a slight 
decrease in stress as the first plastic defor 
mation began. The extent of stress decrease 
in this region is illustrated by the behavior 
of specimen No. 35-1. The stress-strain 
curve for this specimen was linear until 
. yielding occurred at a maximum resolved 
stress of 1465 grams per square millimeter. 
There was a distinct yield point and the 
resolved stress decreased to 1410 grams per 
sq. mm. during the first plastic extension. 
This decrease was soon followed by an 
increase in load as the specimen strain- 
hardened. The yield points or critical maxi- 



mum resolved shear stresses obtained in 
these short-time tests are in agreement 
with the results obtained for similar brass 
crystals by Sachs and his co-workers. 14 - 15 

Those specimens loaded by the weight- 
and-lever method naturally could show no 
decrease in stress as the specimens yielded. 
A stress-strain curve typical of this means 
of loading is illustrated by that for speci 
men No. 8-4 in Fig. i. Stress is proportional 
to strain substantially to the critical re 
solved shear stress, as with other specimens 
tested. The departure from this first slope 
of the stress-strain curve is very great as 
plastic extension begins. 

Creep tests or tests under prolonged 
loading conditions were made with three 
specimens. Specimen No. 12-3 was suc 
cessively loaded at 750, 900, and 1000 
grams per sq. mm. for periods of 505, 335, 
and 335 hr., respectively. No creep could 
be observed in these conditions and, indeed, 



FIG. 2. CREEP-TIME RELATIONS FOR SPECI 
MEN 12-3 AT ROOM TEMPERATURE; MAXIMUM 
RESOLVED SHEAR STRESS INCREASED SUCCES 
SIVELY AS INDICATED. 

the stress-strain curve from zero stress to 
1000 grams per sq. mm. was definitely a 
straight line. The stress-strain curve was 
still linear when the load was increased to 
noo grams per sq. mm. resolved shear 
stress, but creep then occurred at an in 
creasing rate for a period of two weeks, at 
which time the test was terminated. The 
course of the creep-time curve is shown in 
Fig. 2 ; the total creep extension during r this 
period was 0.26 per cent. 



Main 



H. L. BURGHOFF AND C. H. MATHEWSON 



49 



This test appeared to show the existence 
of a definite creep limit between 1000 and 
1 100 grams per sq. mm. resolved stress, 
below which no creep occurred, at least 
within the limit of accuracy of the test, 
The critical stress, elastic limit, or yield 
point, however named, is thus definitely 
lower for prolonged than for brief loading. 
This test gave the first indication that the 
early part of the creep process in these 
single crystals of brass takes place at an 
increasing rate, a phenomenon that was 
verified with other specimens. 

Specimen No. 32-2 was loaded to 1200 
grams per sq. mm. resolved stress, and 
stress was proportional to strain over this 
range. It was intended that creep observa 
tions should be made at this stress and the 
load was accordingly held constant. No 
discernible creep took place in 20 min. 
However, in the ensuing 22-hr, period, 
during which no readings could be taken, 
the specimen had elongated practically 
10 per cent and was still deforming or 
creeping, but at a very low rate, after that 
time. The deformation soon stopped com 
pletely, as no further creep was detected 
in the period from 50 to 137 hr. total time. 
The extension must have started at an 
increasing rate, soon attaining a maximum 
rate and thereafter decreasing to a lower 
rate as the specimen strain-hardened. 

The load was then increased to 1300 
grams per sq. mm. resolved stress; exten 
sion, for which complete data were ob 
tained, began again immediately. The creep 
occurred at an initially increasing rate 
(Fig. 3) and attained a maximum velocity 
within an hour. This rate was maintained 
for a short time but thereafter decreased 
rapidly, so that the creep curve was virtu 
ally flat in 5 or 6 hr. The specimen had 
elongated 2.8 per cent additionally during 
this period and crept only 0.02 per cent 
further from this point through 24 hr. 
elapsed time at this stress. No measurable 
creep occurred from 24 to 193 hr. The re 
sults attained on this specimen were similar 



in character to the results obtained by 
Mehl and Gensamer in creep tests of iron 
crystals. 

Specimen No. 28-1 was also made the 




J00 400 

77f7J ff mfrMfo* 

FIG. 3. CREEP-TIME CURVE FOB. SPECI 
MEN 32-2 AT ROOM TEMPERATURE; MAXIMUM 
RESOLVED SHEAR STRESS = 1300 GM./SQ. MM. 
PREVIOUSLY EXTENDED 10 PER CENT AT 
RESOLVED STRESS OE I2OO GM./SQ. MM. 

subject of a creep test, maximum resolved 
shear stress in this case being 1100 grams 
per sq. mm. The stress-strain curve ob 
tained in loading was linear to this stress. 
As shown in Fig. 4, creep occurred at an 
initially increasing rate, attaining a maxi 
mum velocity between 400 and 700 hr. 
elapsed time, during which the specimen 
extended from 0.5 to 4 per cent. The creep 
rate decreased markedly beyond 700 hr., 
becoming extremely low after 1200 hr. The 
test was finally terminated after about 
1600 hr., at which time the total extension 
was 5.0 per cent. 

The load was then released to 74 grams 
per sq. mm. resolved stress, which low 
value was desirable in maintaining proper 
alignment for observation so that a study 
of creep recovery could be made. The 
specimen immediately contracted 0.034 per 
cent, which corresponded exactly to the 
extension of the specimen when loaded from 
74 to 1 100 grams per sq. mm. No further 
contraction or creep recovery was observed 
over a period of 92 hours. 

During the course of the creep test on 
this specimen, an effort was made to count 
the number of slip lines or bands as the 



1052183 



TIME AND TEMPERATURE IN DEFORMATION OF BRASS CRYSTALS 



test proceeded. This could be done roughly 
in the initial stages but it soon became 
impossible to keep any reasonable record of 
the markings. The slip bands increased in 



velocity, but to a variation in the number 
of operating slip planes. 

As explanation for this, it is visualized 
that slip first occurs on a few planes and 



400 600 

~ 



7//n* srourj 

KIG. 4 CREEP-TIME CURVE FOR SPECIMEN 28-1 AT ROOM TEMPERATURE; MAXIMUM RESOLVED 

SHEAR STRESS == IIOO GM./SQ. MM. 



number at an increasing rate (Fig. 5) until 
about 75 bands had been counted. There 
after slip markings became so numerous 
that any attempt to count them was futile. 



VJ 

*400< 



I 



FIG. 5. RELATIONSHIPS AMONG AMOUNT OF 
CREEP, NUMBER OF VISIBLE SLIP BANDS, AND 
TIME DURING THE INITIAL STAGE OF CREEP OF 
SPECIMEN 28-1. 

It appeared that the number of visible slip 
markings increased directly as extension 
increased and that the variation in creep 
rate was due, not to a change in slip 



then takes place on additional planes 
through a migration of stress concentra 
tions. The initially increasing rate of exten 
sion is viewed as resulting from the great 
availability of unhardened potential slip 
planes with stress concentrations serving to 
accelerate the number of participating slip 
planes during this time. As the metal 
begins to strain-harden and the number of 
available unhardened slip planes decreases, 
the over-all rate of extension naturally de 
creases and may approach zero. This 
explanation may be further amplified by 
considering that, for a given load, slip 
along a participating plane takes place at 
substantially constant rate with no varia 
tion in such individual slip velocity being 
required to explain apparent creep rates of 
a specimen. However, it may well be that 
such slip velocity along a plane is a func 
tion of the applied stress, or additionally, 
the shear or displacement along the plane 
may be a function of the stress. 



H. L. BURGHOFF AND C. H. MATHEWSON 



Such an explanation is not made in terms 
of any particular mechanism of slip and 
contains an assumption that all of the slip 
bands observed may be operating at the 
same time. One of the authors has proposed 
(private distribution, typescript) a form of 
slip in which moments of reversed shearing 
force derived from a rearrangement of the 
atoms at the yield point oppose smaller 
moments of externally applied elastic shear 
ing force, to give slip with a residue of stress 
in strained spacings at and near the site of 
slip. Strain-hardening from this point of 
view is residual stress developed by slip. 
Slow recovery from this condition by 
equalization of internal strain through the 
lattice would decrease the additional stress 
required to produce further slip and remove 
preexisting restrictions as to the site of slip. 
Under constant application of a fixed load 
as in these tests the effect would seem to be 
not unlike that of decreasing friction at a 
sliding contact, thus permitting a quicker 
response to the load. Recovery, however 
visualized, is known to be effective during a 
slow straining process. If complete, no 
strain-hardening occurs and if incomplete 
the normal amount of strain-hardening is 
reduced. The stress-strain curve plotted in 
Fig. i reveals inconsequential strain-hard 
ening in the early stage of deformation 
conducted at a loading rate of approxi 
mately 30 grams per sq. mm. per minute. 

All specimens tested at room tempera 
ture, whether in the short or prolonged 
tests, showed slip markings due to plastic 
extension. In every case it was found that 
these markings corresponded to the {in} 
plane of maximum resolved shear stress. 
For specimens that suffered appreciable 
extension, the change in orientation was as 
prescribed by movement along a {in} 
plane in [101] direction. 

Tests at 300^. 

Both short and long-time tests were 
made at 3ooF., four specimens in all being 
investigated at this temperature. Data con 



cerning the specimens used and some test 
results obtained are summarized in Table i . 
Tests at this temperature indicated that 
the critical shear stress, the stress at which 



O/ 02 0.J 

Axtcr/ S?ra/n - per cent 

FIG, 6. RESOLVED STRESS-STRAIN CURVES 
FOR SPECIMENS 8-3, 1 1-3, AND 30-3 IN TENSION 
AT 3ooF. (CURVES OFFSET FROM ORIGIN FOR 

CLEARNESS.) 

substantial yielding occurs, is somewhat 
lower in short-time tests than the values 
obtained at room temperature. Further 
more, the yielding was much more abrupt 
and was difficult if not impossible to follow 
with the available measuring means during 
the first part of the plastic extension. As 
shown in Fig. 6, the stress-strain curves for 
these short-time tests were substantially 
linear up to the point of actual plastic flow. 

Specimen No. 30-2 was subjected succes 
sively to stresses of 800, 1000, and 1100 
grams per sq. mm. for periods of 95, 70, and 
95 hr., respectively, during which no creep 
was observed. When this stress was in 
creased to 1200 grams per sq. mm., not 
more than 0.0013 per cent extension or 
creep occurred in the first 3^ hr. at the 
higher stress, but before 28 hr. had elapsed 
(during which time no intermediate read 
ings could be taken) the specimen had 
extended 10 per cent. It must be assumed 
that such extension took place with initially 
accelerating and later decelerating rate. No 
further creep was observed after this time 
up to 1 68 hr., when the test was terminated. 

Specimen 30-3 was held for 96 hr. at 1000 
gram per sq. mm. resolved stress and no 



TIME AND TEMPERATURE IN DEFORMATION OF BRASS CRYSTALS 



apparent creep occurred. When the stress 
was increased to 1 100 grams per sq. mm. no 
apparent plastic extension occurred^for a 
minute or so and then a barely perceptible 







O/ 0.2 0-3 04 0S 

Axial Strain - per cent 

FIG. 7. RESOLVED STRESS-STRAIN CURVES 

FOR SPECIMENS 14-! , 11-4, 30-!, AND 2p-2 
IN TENSION AT 5OOF. 

movement or deformation of the specimen 
was noticed, which increased in rate, then 
decreased, and stopped. The extension, 
which was nearly 7.5 per cent, took place in 
an interval of not more than 2 min. This 
test was continued for a period of 163 hr., 
during which time no further extension 
could be observed. There was again indi 
cated a time-elongation curve of shape 
similar to those obtained at room tempera 
ture with rate first increasing, later de 
creasing, and finally approaching zero as 
hardening occurred. 

Slip markings were visible on all speci 
mens tested at this temperature and corre 
sponded to the {in} plane of maximum 
shear stress. Orientation changes were as 
usually prescribed in terms of slip and 
rotation. 

Tests at 500?. 

Both short-time and long-time tests 
were run at 5ooF., four specimens being 
thus tested. Specimen data and some test 
results for this group are also given in 
Table i. 

The short-time tests showed, as at 
3ooF., that yielding, once started, is much 



more abrupt than at room temperature and 
continues for a considerable extension 
without increase in load. The critical re 
solved shear stress was found to be only 
slightly less than at room temperature. The 
tests made for conditions of prolonged 
loading indicated the existence of a creep 
limit essentially the same as at room 
temperature; i.e., approximately 1000 
grams per sq. mm. maximum resolved shear 
stress, in spite of a temperature well above 
the lowest possible recrystallization tem 
perature of polycrystalline metal of the 
same composition. 

Specimen No. 30-1 was successively 
loaded to 300, 400, 600, 800, and 1000 
grams per sq. mm. resolved stress for 
periods of 165, 409, 166, 168, and 72 hr., 
respectively. At none of these stresses was 
creep observed. The stress-strain data 
obtained in the interrupted loading of the 
specimen over this long period of time show 
(Fig. 7) a substantially straight line or 
proportional range up to 1000 grams per sq. 
mm. and also up to 1100 grams per sq. mm., 
which was the highest and final stress used. 
When the load was increased to this last 
value, no visible extension of the specimen 
occurred during at least 6 hr. However, 
after that time and in less than 48 hr., the 
specimen had elongated 4 per cent and then 
the extension had presumably been stopped 
because the loading lever came to rest 
against the supporting structure. It would 
appear, therefore, that the first extension 
must have taken place at an increasing 
rate. 

It was evident that within the limits of 
the test there was a creep limit for this 
specimen between 1000 and noo grams per 
sq. mm. resolved shear stress. 

Stress-strain data were obtained for 
specimen No. 29-2 up to 1000 grams per sq. 
mm. resolved stress. The stress-strain curve 
was linear (Fig. 7) up to 935 grams per sq. 
mm., but deviated slightly in the last load 
ing increment. The specimen was loaded 
immediately above the short-time propor- 



H. L. BURGHOIT AND C. H. MATHEWSON 



53 



tional limit for the prolonged test that was 
then started. There was no visible creep for 
i J^ hr. and only a possible slight indication 
of creep (0.00135 per cent) after 2 hr. How 
ever, in 22 hr. the creep mounted to 0.0155 
per cent, after which the rate decreased and 
became almost negligible after 168 hr., the 
total time at this stress. The now familiar 
creep curve with initially increasing and 
later decreasing rate therefore describes the 
deformation at this stress. 

Maximum resolved shear stress was then 
increased to noo grams per sq. mm. and no 
visible creep occurred for almost 2 hr. Then 
while a reading for the 2-hr, interval was 
about to be taken the specimen began to 
yield or slip visibly and uniformly for not 
longer than i min., when apparently it 
stopped. It was established that the speci 
men had elongated 2.5 per cent during this 
interval. It was obvious that the sudden 
deformation had started at an increasing 
rate and later had slowed down. The same 
load was maintained for a total time of 263 
hr. after the sudden 2.5 per cent extension 
and it was found that there was a rapid 
although small flow to 0.023 per cent addi 
tional extension in 3 hr. after the sudden 
yielding. After this, creep rate was much 
less and tended to decrease with increasing 
time. 

Load was successively increased to 1150 
and 1230 grams per sq. mm. for periods of 
239 and 167 hr. Creep occurred at a de 
creasing rate for both stresses and both 
rates were little greater than the final rate 
at noo grams per square millimeter. 

When the load was finally increased to 
1435 grams per sq. mm. resolved stress, the 
specimen immediately yielded and its 
movement was stopped only when the lever 
came to rest on the supporting frame. 

Slip markings were observed on all speci 
mens tested at this temperature. These 
markings again corresponded to the {in} 
plane of maximum resolved shear stress and 
orientation changes were of the normal 
type. 



Tests at 700^. 

The tests at 7ooF. yielded results among 
the most interesting of the entire investiga 
tion, as it was more readily possible to 



0J 

Axial Strain -per cenf 

FIG. 8. RESOLVED STRESS- STRAIN CURVES 
FOR SPECIMENS I AND 12-4 IN TENSION AT 

7 00F. 

follow the course of creep than at 300 or 
5ooF. Moreover, greater differences in 
behavior were found between short and 
prolonged tests. Data for the test speci 
mens are given in Table i. 

Stress-strain curves obtained for the 
ordinary rate of loading of specimens i and 
12-4 are shown in Fig. 8. The curves show 
abrupt and considerable yielding at maxi 
mum resolved shear stresses of 1290 and 
1420 grams per sq. mm., respectively, values 
that are approximately the same as those 
obtained in the previously described room- 
temperature tests. 

Slip markings were visible on the surface 
of these specimens in spite of oxidation, and 
were found to correspond to the proper slip 
plane as predicted for room-temperature 
deformation. 

Creep data were obtained for specimen 
No. 7-4 for maximum resolved shear 
stresses of 200, 300, and 450 grams per sq. 
mm., respectively. In spite of being far 
below the critical stress or yield point of the 
short-time test, each of these stresses 
produced distinct creep at ever increasing 
rates (Fig. 9). The testing of this specimen 
was terminated when the total creep for the 



54 



TIME AND TEMPERATURE IN DEFORMATION OF BRASS CRYSTALS 



entire test was 0.22 per cent. The deforma 
tion being small and the surface oxidized, 
no slip markings were visible. 
It seems logical to assume that creep 




FIG. 9. CREEP-TIME CURVES FOR SPECI 
MEN 7-4 AT 700F.; MAXIMUM RESOLVED 
SHEAR STRESSES INCREASED SUCCESSIVELY AS 
INDICATED. 

would have taken place at any stress at this 
and higher temperatures. Furthermore, it 
would seem probable that higher stresses in 
the vicinity of the knee of the stress-strain 



amounted to 0.125 per cent. The curve 
thereafter became concave to the time axis 
as creep rate decreased. 

No slip markings were visible on the 
blackened surface of the specimen, but it 
was found that the cross section had 
changed definitely, although only very 
slightly, from a circular to an elliptical 
shape, which made possible the location by 
X-ray means of the zone containing the 
operating slip plane. This zone proved to 
contain the predicted { 1 1 1 } plane of maxi 
mum resolved shear stress. There was again 
no indication of any new or different 
manner of deformation. 

SUMMARY 

As tested by the usual tension-testing 
procedure, at room temperature, single 
crystals of 71-29 brass show a small but 
definite drop in load as plastic deformation 
first begins, indicating the existence of a 
yield point as in iron crystals. Critical maxi 
mum resolved shear stress in slow tensile 



^ 

*0.4 



FIG. 10. CREEP-TIME CURVE TOR SPECIMEN 31-3 AT 7ooF.; MAXIMUM RESOLVED SHEAR STRESS : 

600 GM./SQ. MM. 



curve would give results similar to those for 
specimen No. 29-2 at 5ooF., but with 
greater creep and less hardening effect. 

Specimen No. 31-3 was subjected to 
prolonged loading equal to a maximum re 
solved shear stress of 600 grams per sq. mm. 
for 625 hr. The course of the creep curve is 
shown in Fig. 10. Creep rate increased dur 
ing the first 100 hr., at which time extension 



tests ranged from about 1300 to 1465 grams 
per sq. mm. at room temperature, agreeing 
well with test results of previous investiga 
tions, and was not substantially changed at 
300, 500, and 7ooF. The yielding or first 
plastic deformation in such tests became 
more abrupt and rate of the yielding be 
came more rapid as the test temperature 
increased. 



DISCUSSION 



55 



Creep tests or tests under conditions of 
prolonged, constant load revealed the 
existence of a true elastic limit or creep 
limit, at least within the sensitivity of the 
tests, below which no plastic deformation 
was observed throughout periods as high as 
several hundred hours. This limit corre 
sponded to a maximum resolved shear 
stress of about 1000 to noo grams per sq. 
mm., and was substantially unchanged 
from room temperature to 5ooF. At 
7ooF., appreciable creep occurred for 
resolved stresses as low as 200 grams per 
sq. mm. and it seemed probable that creep 
would take place at any stress at this or 
higher temperatures. 

The initial plastic deformation in the 
creep tests at all temperatures took place at 
an increasing rate, the rate later decreasing 
and, for the lower temperatures, apparently 
approaching zero. Evidence was found to 
show that the amount of creep was directly 
proportional to the number of slip bands 
defining planes that must have operated 
either simultaneously or in succession dur 
ing the process. Alternative explanations of 
this structural effect are possible. 

Observation of dimensional changes and 
slip markings correlated with X-ray analy 
sis indicated that the manner of deforma 
tion was the same for all tests in the 
investigation; that is, by slip upon {in} 
planes in [101] directions with orientation 
changes as classically predicted. 



REFERENCES 

C. L. Clark and A. E. White: Proc. Amer. Soc. 
Test. Mat. (1932) 32 (II), 492. 

D. Hanson and M. A. Wheeler: Jnl. Inst. Metals 
(1931) 45, 229. 

B. B. Betty: Proc. Amer. Soc. Test. Mat. (i93S) 
35 (ID, 1-93. 

Baker, Betty and Moore: Trans. A.I.M.E. (1938) 

128, 118. 

R. P. Miller: Trans. A.I.M.E. (1936) 122, 176. 
R. F. Miller and W. E. Milligan: Trans. A.I.M.E. 

(1937) 124, 229. 

M. Gensamer and R. Mehl: Trans. A.I.M.E 

(1938) 131, 372. 

A. Greninger: Trans. A.I.M.E. (1935) 7t 75- 
G. I. Taylor and C. F. Elam: Proc. Roy. Soc. 
(1925) io8-A, 28. 

C. F. Elam: Distortion of Metals. Oxford, 1935- 
Clarendon Press. 

E. Schmid and W. Boas: Kristallplastizitat. 



13. H. L, Burghoff: Trans. A.I.M.E. (1940) 137, 214. 

14. G. Sachs and M. Masima: Ztsch. Physik (1928) 

50, 161. 

15. von Goler and G. Sachs: Ztsch. Physik (1928) 55 

581. 



DISCUSSION 

(/. /. Kanter presiding) 

R. M. BRICK* New Haven, Conn. As 
everyone knows, who has worked with this type 
of cast crystal, the specimens were not true 
single crystals. Laue reflection pictures of the 
crystals usually show multiple spots. The 
angular difference between the different 
dendritic axes varies from one crystal to 
another. I should like to ask the authors if they 
noticed, studying all these crystals with vary 
ing degrees of imperfections, whether there was 
any difference in the critical resolved shear 
stress at a given load and temperature be 
tween a rather imperfect crystal and one less 
imperfect. 

A second point relates to the author s re 
marks in regard to observations of slip markings 
on the surface of the crystals. These are 
interesting, particularly so in regard to claims 
that creep is not essentially a slip mechanism 
but a viscous flow mechanism. The slowing 
up in the creep rate and ultimate stoppage is 
attributed to possible exhaustion of unstrain- 
hardened slip planes. I should like to mention 
some tensile deformation experiments on single 
crystals that Dr. Burghoff made up but did not 
use, and left at the laboratory. 

At a 2 per cent extension (under ordinary 
loading rates) which in some of his experiments 
might represent the maximum extension ob 
tained before creep stopped, numerous slip 
markings would be visible to the naked eye. 
If a flat section of the specimen had been 
polished before the extension occurred, and 
one of these slip lines was examined at a high 
magnification of 1000 diameters, it would be 
found to be made up of a tremendous number 
of lines. We have made tests where the crystal 
was stressed to a very slight extension of 
perhaps o.i per cent, at which point a few 
markings first appeared. In one specific 
"macro" line, five or six of these slip markings 
would be visible at 1000 diameters. On con 
tinued loading of the crystal, the number of 
these markings increased as Dr. Burghoff indi- 



12. Bull. 4 



Berlin, I93S- Julius Springer. 
nil. 410. Nat. Bur. Stds. 



* Hammond Laboratory, Yale University. 



TIME AND TEMPERATURE IN DEFORMATION OF BRASS CRYSTALS 



cated, but, also, at any one single marking 
where previously there were five or six this 
might increase up to 20, 40, or 50; within one 
single location or what would appear to be one 
line to the naked eye. At extensions approach 
ing the maximum mentioned by the authors, 
there were still large areas of apparently com 
pletely undisturbed metal, and it seems rather 
incredible that there is not a very large number 
of undisturbed planes still available for slip. 

M. GENSAMER,* Pittsburgh, Pa. In our 
work to which Dr. Burghoff referred, we 
observed an interesting and puzzling thing. 
When we put on the low loads, there was no 
creep, of course; when we put on a high enough 
load, we obtained the kind of creep that Dr. 
Burghoff described. With a little bit more load, 
nothing happened. We might make another 
slight increase in load with no creep. Finally, 
with a sufficient additional load, creep would be 
observed. This process could be repeated after 
each creep observation; we could put a little 
more load on and not get any more creep, at 
least not in the time that we could afford to 
wait. I should like to know whether Dr. 
Burghoff and Dr. Mathewson have observed 
a similar effect. 

In connection with Dr. Brick s remarks, is 
not this kind of behavior to be expected when 
the materials are not really homogeneous? 
After all, it is expecting a great deal to suppose 
that every part of a crystal will have the same 
yield point, the same creep limit. 

B. B. BETTY,| Huntington, W. Va. I have 
observed some of the things mentioned, in 
my work on lead. I think spacing of the slip 
lines, rather uniform spacing in the course of 
deformation, with partially strained material in 
between, must be accounted for on the basis 
of the observations of more detailed and 
intensive studies that have been made on 
aluminum single crystals in the National 
Physical Laboratory, from which it must be 
concluded that the strain-hardening proceeded 
at right angles to the plane of slip as well as in 
the direction of plane of slip. Undoubtedly 
some of these phenomena are due to the 
involvement of a larger number of planes as 



* Carnegie Institute of Technology. 

t Research Laboratory, International Nickel Co. 



the degree of work-hardening is intensified. A 
convenient assumption to make here is that 
the intensity of work-hardness decreases with 
increasing distance from the plane in which 
slip has occurred. 

I am interested to see this S-curve. since I 
have observed something of that kind in 
polycrystalline specimens for which I have not 
heretofore had an adequate explanation. 

H. L. BUE.GHOFE (author s reply). As Dr. 
Brick says, the crystals were imperfect, as 
shown by the character of the spots of Laue 
reflection pictures, and there were real differ 
ences in critical resolved shear stresses for given 
loading conditions. Although no critical study 
of this situation was made, there was no 
apparent relation between crystal imperfection 
and critical stress. 

With regard to his second point, it is realized 
that the unaided eye cannot see the detail of 
slip markings made evident by a microscope. 
However, it was felt that the reporting of the 
occurrence of the slip markings in the partic 
ular test cited was interesting and of speculative 
value. 

Referring back to Figs. 4 and 5, the slip 
bands that were counted were, of course, only 
the first ones. They increased so rapidly after 
400 hr., being distributed over the entire 
gauge length, that it was no longer feasible to 
count them. The mechanism of hardening is 
naturally involved in these considerations, 
and I think that the theory that Dr. Mathew 
son is now propounding may lead to some light 
on the spacing of slip bands and the appearance 
of subsequent ones. Mr. Betty s remarks are 
interesting in that they also include considera 
tion of the hardening process. 

Dr. Gensamer asked if we had observed any 
effect due to application of additional load 
after the original yielding. This is illustrated 
in the paper for specimen No. 29-2, which was 
tested at 500. Considerable yielding was ob 
tained on application of a certain stress. When 
two or three higher stresses were applied, the 
specimen still crept, of course, but still at 
decreasing rates not greatly different from that 
for the first lower stress. When a still higher 
load was finally applied, a sudden yielding was 
experienced, and the specimen elongated as 
far as the equipment would permit. 



Flow of Solid Metals from the Standpoint of the 
Chemical-rate Theory 



BY WALTER KAUZMANN* 

(New York Meeting, February 1941) 



ALL viscous or plastic flow of incompres 
sible matter is the result of shear strain; the 
changing shape of any body that is being 
plastically deformed can be completely 
described in terms of the shear strain 
occurring at each point in the body.f 
Furthermore, the relative amounts of shear 
strain occurring in different parts of a 
body under stress depend upon the relative 
rates of shear strain at those different 
points, so that if the dependence of the 
shear rate of any material on the tempera 
ture, shear stress, previous history, etc., 
were understood, the principles of plastic 
deformation would be known, and we 
could calculate how a body made of a given 
material, having a given shape, and subject 
to a specified system of forces, would 
change its shape with time. It may be said, 
then, that the problem of plasticity resolves 
itself into, first, the problem of the rates of 
pure shear processes, and second, the 
application of what is known about pure 
shear to actual cases. It is the purpose of 
this work to investigate the former problem 
i.e., the dependence of shear rates on 
temperature, stress, etc. from the stand 
point of recent developments in the field of 



Manuscript received at the office of the Institute 
Dec. 2, 1940. Issued as T.P. 1301 in METALS TECH 
NOLOGY, June 1941. 

* Westinghouse Research Fellow, Westinghouse 
Research Laboratories, East Pittsburgh, Pa. 

f Thus, a plate undergoing deformation as 
in Fig. i may be regarded as made up of a 
large number of tiny squares, and the change 
in the shape of the plate is the result of the 
changed shapes of the squares, which in turn 
are describable as pure shear strains. 



chemical-rate theory and with particulai 
reference to the phenomena occurring in the 
creep of metals and in the plastic flow of 
crystalline solids in general. It is left for 
the mechanical engineer, and perhaps the 
expert in hydrodynamics, to deal with the 
problem of applications. 1 

RESOLUTION OF MACROSCOPIC SHEAR 

INTO MICROSCOPIC MOVEMENTS; 

RESEMBLANCE TO CHEMICAL 

REACTIONS 

Just as any large body is made up of 
many small units, or atoms, so macro* 
scopically observable shear is the result of 
many microscopic unit shear processes; just 
as the understanding of the behavior of 
large bodies demands the understanding of 
the atoms of which it is made, so the under 
standing of macroscopic shear demands the 
understanding of these fundamental unit 
processes. The study of shear from this 
"atomistic" point of view has made con 
siderable progress. There has been fairly 
complete understanding of shear processes 
in gases for a long time, at least for non- 
turbulent flow, as a chapter on gaseous 
viscosity in any text on the kinetic theory 
of gases will prove. In recent years Andrade 2 
and Eyring 3 have presented important 
theories for the flow of liquids and amor 
phous solids. Becker, 4 Orowan, 6 Taylor, 6 
Burgers, 7 and Kanter 8 have considered the 
plastic flow of crystalline solids from this 



i References are at the end of the paper. 



57 



FLOW OF SOLID METALS AND CHEMICAL-RATE THEORY 



point of view, and have derived formulas structures in a body whose motions past 



for the correlation of shear with various 
other properties of the flowing matter, while 
Polanyi, Schmidt, Smekal, and many 




FIG. i. PLASTIC DEFORMATION ARISING FROM 

PURE SHEAR. 

others have constantly striven to interpret 
the results of their work on the deformation 
of crystalline material in terms of the 
molecular process occurring in shear. 

The point of view that has been adopted 
in the present investigation of the shear of 
crystalline solids is that of Eyring, whose 
general theory of shear rates, 3 based upon 
his statistical mechanical equation for rate 
processes is admirably suited to this prob 
lem. 9 Since, however; this approach was 
developed with a view to its use in the study 
of the rates of chemical reactions, the con 
cepts it involves are those most familiar to 
the chemist, while, on the other hand, the 
flow of crystalline solids has been of interest 
chiefly to the engineer and physicist, who 
are less familiar with these attitudes of the 
chemist. Therefore, it seems desirable to 
review here at some length the notions 
behind the rate equation of Eyring and its 
application to the general problem of shear 
rates, and only then to apply it to the 
specific problem of shear rates in crystals. 
Before doing this, however, it will be 
profitable to introduce a few simple notions 
that will serve to make clear just what 
features of the problem are similar to those 
characteristic of chemical reactions. 

In what follows, use will be made of the 
concept of units of flow; these are the 



one another make up the unit shear process 
mentioned above. In a liquid such as water 
the units of flow are believed to be single 





a b c 

FIG. 2. MOLECULAR SHEAR PROCESS IN 

ORDINARY LIQUIDS. 

molecules, and the unit process is the 
passage of single molecules past one another 
in some such manner as is illustrated in 
Fig. 2. Here the motion of the molecules 
through the stages a, b, c has resulted in a 
shear strain of \/L in the direction indi 
cated by the solid arrows, while the rate of 
shear is equal to the shear resulting from 
each jump times the number of jumps 
occurring in unit time. Clearly we can in 
general say that 

, X 

shear rate = 5 = j v [i] 

where X is the average distance in the shear 
direction moved by units of flow past one 
another in one jump, L is the average 
distance between layers of units of flow, and 
v is the net number of jumps per second 
made by the unit of flow in the shear 
direction. 

If we can understand how the quantities 
X, L, and v depend on the shear force, 
the temperature, etc., we shall have as 
complete an understanding of the phe 
nomenon of shear as will be required for 
application by the engineer to problems of 
plastic flow in general. 

In solids and liquids, to which the view 
point of equation i is particularly well 
suited, the distance moved by units of flow 
and the distance between layers of units of 
flow are expected to be fixed chiefly by the 
lattice dimensions of the flowing materials. 



WALTER KAUZMANN 



59 



The dependence of these on the shear force, 
temperature, cold-working, and other fac 
tors, would certainly not be sufficient alone 
to account for the very large changes in 
shear rates that such factors are known to 
be capable of causing. This makes impera- 



exponential dependence of the rate of most 
chemical reaction on the temperature as 
follows : 

rate = Ce* H * /RT [2] 

where A#* is the activation energy re- 



POTENTtAL 
ENERGY 



ACTIVATED 
REACTANT STATE 



PRODUCT 




AH* FOR B--A 



COURSE OF REACTION 

FIG. 3. TYPICAL POTENTIAL-ENERGY CURVE FOR REACTION B 

ENERGY. 



* A, INVOLVING AN ACTIVATION 



tive the thorough understanding of the 
term involving the rate at which units of 
now pass one another, since this term 
probably contains the greater part of the 
sensitivity of shear rates to the external 
conditions. 

It is very likely that in order for units of 
flow to be able to pass one another, espe 
cially in solids, a considerable amount of 
energy is required. (Thus, in Fig. 26, if the 
surrounding molecules were shown, it 
would be apparent that there is a great deal 
of squeezing of the molecules against one 
another.) This energy may be considerably 
greater than the ordinary thermal energy of 
the molecules, so that in order for shear to 
occur between two units of flow, these must 
be activated. The number of passages of 
units of flow per second, therefore, is very 
greatly reduced, since only relatively few of 
them possess much more than the average 
thermal energy. 

This is exactly the situation that faces 
the chemist when he seeks to understand 
the rates at which substances react. Here, 
too, the reacting molecule or molecules 
must be activated before the reaction can 
proceed. As is well known, this requirement 
of an activation energy gives rise to an 



quired by the reacting molecules before 
they can react, R is the gas constant 
(= 1.987 cal. per mol per degree centi 
grade), T is the absolute temperature and 
C is a constant independent of the tempera 
ture. It will be found very helpful to repre 
sent the course of a reaction by an " energy 
map"; a typical, simplified one is shown in 
Fig. 3. The ordinate of such a plot repre 
sents the potential energy at any point 
along an abscissa (here given as "course of 
reaction") which usually involves one or 
more interatomic distances. Then if the 
reaction in question is B > A , the AH* of 
Eq. 2 is the same as that shown in Fig. 3, 
while if the reaction is A -+B, the AZI* of 
Eq. 2 is A#* + AH in Fig. 3, AH being 
the heat of Ike reaction. 

Apparently a relationship similar to Eq. 
2 is required in order to obtain the impor 
tant quantity v in Eq. i. The form of this 
relationship can be derived if the unit shear 
process (passing of units of flow) is regarded 
as a chemical reaction, and if the equations 
well known to chemists are applied in 
connection with the rates of chemical 
reactions. It will therefore be appropriate 
at this point to examine these equations in 
some detail. 



6o 



FLOW OF SOLID METALS AND CHEMICAL-RATE THEORY 



CHEMICAL THERMODYNAMICS AND THE 
RATES OF CHEMICAL REACTIONS 

From thermodynamics we know that for 
a system in equilibrium e.g., A ^ B the 
relative amounts of A and B present are 
given by 



(A) 



\ M 

L3J 



where (A) and (B) represent the amounts 
of A and B present, and AF is the change in 
free energy resulting from the change of one 
mol of A into one of B. Furthermore, 



AF = AH - TAS 



[4] 



where AH is the change in ordinary energy 
(heat of reaction) and AS is the change in 
entropy resulting when the same change 
occurs. We can thus write 



(2 



[Si 



The true significance of this phenome- 
nological description of an equilibrium in 
terms of the factors e* s/R and e-^ /RT is 
made possible by statistical mechanics, and 
that very simply and directly. The term 
Q-AH/RT i s a consequence of the well-known 
Boltzmann relation, which states that when 
anybody is considered as made up of a large 
number of particles having a definite total 
energy and an absolute temperature 7 1 , the 
relative probability that any particle is in 
an accessible state* having an energy e is 
given by e~ e/kT where k is a universal con 
stant known as the Boltzmann const an t.f 
Then if the average energy of the molecules 
of B is higher by Ae per molecule than that 
of the molecules of A , and if the states of B 
are accessible to those of A } the relative 
amounts of B and A present must be pro 
portional to e-^ kT . Since H = NAe and 
R = Nk, where N is the number of mole 
cules in a mol (Avogadro s number), this 

* The significance of the term "accessible" 
will be brought out below. 

t For derivations of the Boltzmann equation, 
see references 10, n and 12. 



factor can be written as e~ AH/RT , as given in 
Eq. 5- 

The significance of the term c* s/R , on the 
other hand, can best be brought out by an 
example. Suppose the state A corresponds 
to molecules of a perfect gas moving in a 
box at sea level and that the state B corre 
sponds to the molecules in a second box 
at an altitude of h centimeters above sea 
level and having the same volume as the 
first box. Let the two boxes be connected by 
means of- a pipe so that the molecules can 
easily get from one box to the other. (The 
significance of the term "accessible" 
should now be clear.) Then, according to 
the Boltzmann equation, the ratio of the 
number of molecules in B to that in A is 
given by e~ A * /ftr , where A, the difference 
in energy of a molecule in A and one in B, 
is equal to mgk, m being the mass of a 
molecule and g being the acceleration due 
to gravity. Thermodynamically, 



(A) 



[6] 



and for the reaction, one mol of gas in 
box A > one mol of gas in box B, we see 
from Eq. 3 that AF = Nmgh = Mgh per 
mol, where M Nm = the molecular 
weight of the gas. Since this obviously 
corresponds entirely to a change in energy, 
AF AH = Mgh and AS = o. Now sup 
pose the upper box has a volume three 
times that of the lower one instead of equal 
to it. We must, then, multiply the relative 
probability of B by 3, since each molecule 
now has three times as much space, or 
freedom, in B as in A. Thus, 



(A) 



[7] 



and 



AF = Mgh - T(i.oggR) 
AS = I.O99-R 
AH = Mgh 

On the other hand, if B had only one-third 
the volume of A, the result would be 
AH = Mgh and AS = i 



WALTER KAUZMANN 



6l 



Hence the entropy change in a reaction 
is a measure of what may be called the 
relative freedoms of the reactant and 
product of the reaction: if AS is positive, 
the product has more freedom than the 
reactant, while if it is negative, there are 
restrictions imposed on the product that 
are not imposed on the reactant. Further 
more, the actual numerical ratio of these 
restrictions is given by e^ s/R . 

A striking illustration of the importance 
of the change in the entropy in determining 
the conditions of equilibrium is found in 
the denaturation reactions of proteins 
(the kind of reaction that occurs when an 
egg is boiled or when a man is shot dead). 
Here, entropy increases of the order of 
100 cal. per mol per deg. C. are observed, 
which corresponds to e^ s/R = io 22 . That 
is, the denatured protein molecule can do 
something like io 22 as many things as the 
undenatured molecule. This tends to make 
the equilibrium shift in favor of the de 
natured molecules at very low tempera 
tures. On the other hand, AH for these 
reactions is also large the denatured 
molecule being less stable by something of 
the order of 100,000 cal. per mol and this 
tends to make the equilibrium shift only 
at very high temperatures. The result of 
the operation of these two factors in 
opposite directions is that the equilibrium 
shifts from one side to the other very 
rapidly at ordinary temperatures often 
within a range of a few degrees centigrade. 
It also makes the equilibrium very sensitive 
to small changes in other external condi 
tions, and probably causes much of the 
great complexity shown in the behavior of 
living matter. It is evident, then, that the 
large entropy and heat changes occurring 
in protein denaturation reactions can fur 
nish valuable clues to the answer to the 
question so vital to everyone of what 
happens during denaturation.* 



* Although this is hardly the place for a 
discussion of this problem, it might be men 
tioned that the best interpretation of these 



Large entropy effects are encountered in 
the flow of polymers 13 and, as will be shown 
later, in the flow of solid metals; here they 
furnish important clues concerning the 
mechanism of the processes with which they 
are associated. 

All that has been said in the preceding 
pages has referred to the properties of mat 
ter in equilibrium. This, of course, is the 
limitation that always is imposed on ther 
modynamics, but the more intimate view 
of the behavior of matter provided by 
statistical mechanics has been utilized by 
Eyring to make possible a thermodynami- 
cal treatment of molecular processes occur 
ring at a constant rate and hence not 
restricted to matter in equilibrium. 14 15 

Looking at the "energy map" of a typi 
cal chemical reaction (Fig. 3), we may say 
that the rate at which reactants go to form 
products (i.e., the rate of the reaction) is 
given by KN, where N is the number of 
molecules which in one second move in the 
direction corresponding to reaction past 
the point marked activated state (i.e., past 
the highest point on the easiest path from 
reactant to product) and K represents the 
fractions of these molecules that ultimately 
become product molecules. Unless the 
topography of the energy map is compli 
cated by tortuous weavings or certain 
other features, none of which are at all 
likely to be encountered here, K is very 
close to unity, so can be neglected. 

Eyring has calculated N by regarding a 
molecule in the activated state (an " acti 
vated complex," as he calls it) as nothing 
more than an ordinary molecule, save only 
that this molecule behaves strangely along 
the normal coordinate of its motion corre 
sponding to its "reaction coordinate" (i.e., 

large numbers is that a protein molecule, when 
undenatured, probably has a very regular and 
definite structure. By breaking a few of the 
bonds suitably located in this structure (this 
requiring a lot of energy), the regular structure 
can give way to a myriad of possible structures 
whose very great number accounts for the large 
entropy increase. 



62 



PLOW OF SOLID METALS AND CHEMICAL-RATE THEORY 



the coordinate, motion along which corre- inverse of the time for a single molecule 



sponds to reaction; the abscissa of Fig. 3). 
Then N is the number of activated com 
plexes multiplied by the rate at which each 



to react. 

The value of this formulation of reaction 
rates lies in the fact that AF* is subject to 



POTENTIAL 
ENERGY 




FIG. 4. POTENTIAL-ENERGY CURVE FOR UNIT SHEAR PROCESS. 



passes the point P in Fig. 3 moving in the 
direction corresponding to reaction. But as 
long as the activated complex is in equilib 
rium with the reactants A, thermodynamics 
can be applied, and it will be true that 

Number of activated complexes _ 
Number of reactant molecules ~~ 

[8] 

where AF is the free energy of activation of 
a mol of reactant molecules. This AF in 
cludes the properties of the activated com 
plex along the reaction coordinate; if the 
part of AF depending on these properties 
is separated, and if a more or less straight 
forward statistical mechanical calculation 
of the rate of passage of molecules past P 
is carried out, it is found that 



[9] 



where AF* is now the free energy difference 
per mol between the reactant and the 
activated complex when the latter is re 
garded as an ordinary molecule, except 
that the degree of freedom corresponding 
to the reaction coordinate has been dis 
regarded, it having been absorbed in the 
factor kT/h. For our purposes, Eq. 9 gives 
the rate of any reaction in units of the 



the same sort of interpretation for reaction- 
rate theory as AF was for equilibrium 
theory. It can be broken up into A#* and 
AS*, the heat and entropy of activation, 
each of which can tell a great deal about 
reaction mechanisms, just as AF and A3 1 
tell a great deal about what happens in 
many equilibria (e.g.,. in the denaturation 
of the proteins, as shown above). 

APPLICATION OF EQUATION 9 TO THE 
PROBLEM OF SHEAR RATES 

We are now ready to return to the prob 
lem proposed on page 59 to calculate the 
number of jumps in a given direction that a 
unit of flow can make in one second. Sup 
pose the energy map for a unit shear process 
is that shown as EQ(<$) of Fig. 4. Here, d is 
the distance in the shear direction moved 
by units with respect to one another. The 
rate of passage of units past one another in 
either direction will then be 



[10] 



where AF* is the difference in free energy of 
the units of flow in the normal and acti 
vated states. (If k is in ergs per degree and h 
is in erg-seconds, n is the number of 
passages per second of one unit past other 



WALTER KAUZMANN 



units.) In the absence of applied shear 
forces, the number of jumps, n+, in the posi 
tive d direction (that is, resulting in shear 
in the same direction as that shown by the 
arrows in Fig. 2) will be the same as that, 
W-, in the negative d direction, and there 
will be no net shear. But if a shear stress/ 
per unit area is applied in the positive d 
direction, the potential energy curve 
Eo(d) of Fig. 4 will be modified, taking on 
instead the shape of the dotted line, with 
the form E Q (d) Adf, where A is the 
projected area of each unit of flow in the 
shear plane (i.e., the area on which/ acts, so 
that Af is the actual force acting on each 
unit of flow, and Adf is the linear potential 
field due to the shear force). 

The number of jumps of units of flow per 
second in the positive direction with respect 
to one another will now no longer equal the 
number in the negative direction, since the 
activation energies for the two directions 
are no longer equal. Instead, the activation 
energy for motion in the positive direction 
will be lowered by an amount Alf per 
molecule, while that in the other direction 
will be raised by the same* amount (/ being 
the distance through which the shear stress 
acts in carrying the unit of flow from the 
normal to the activated state). 
Then 

kT 



Substituting in Eq. i, 



and 



J?T 

_ 

k 



[12] 



and v, the net number of jumps in the plus 
direction per unit of flow per second is 



= n + - n. = 



kT 
_ 

h 



-AF*/RT 



_ aih(A.lf/kT) [14] 

* This is only approximately true, and that 
for small stresses. See p. 64. 



2\kT 



shear rate = s = y , * - -* --- 

sinh (Alf/kT) -[15] 

(Note that ordinarily, L = 2l.) This is the 
general equation for shear developed by 
Eyring. 

If Alf <g kT (that is, if the energy sup 
plied by the action of the shear force in 
aiding a unit of flow is small compared with 
its thermal energy), since sinh x ^x for 
small values of x, 



[16] 



That is, the shear rate is proportional to the 
shear stress. This constitutes ordinary 
viscous or Newtonian flow, such as is 
exhibited by water, and the coefficient of 
viscosity is given by 

/ Lh 
^ """ s ~ 2\IA e e l ^ 1 ^ 

This equation has been applied by Eyring 
and co-workers to the flow of ordinary 
liquids, with many interesting results. 13 16 " 19 
If, on the other hand, Alf^> kT (if the 
energy supplied by the shear force in 
activating the molecule is much larger than 
the thermal energy; also, if for any reason* 
w_has a much smaller value than that given 
in Eq. 12), 



which leads to the well-known law for the 
creep rates of metals: 

log s = C + Df [19] 

C and D being constants at any given 
temperature and for any given materiaLf 



* See Discussion of Theory. 

f The hyperbolic sine law for creep rates 
has been suggested previously on phenomeno- 
logical grounds (e.g., by Nadai, in the S. Timo- 
shenko 6oth Anniversary Volume, p. 165), but 
there is little available experimental evidence 
for testing it. It is apparent that for substances 



6 4 



FLOW OF SOLID METALS AND CHEMICAL-RATE THEORY 



DISCUSSION or THE THEORY 

Before applying Eq. 15 to some actual experi 
mental data, it will be well to investigate some 
of the possible difficulties that might arise from 
such an application. We may say at the outset 
that the theory is particularly designed for the 
interpretation of shear processes that require 




Y IGf 5. POTENTIAL-ENERGY CURVES FOR UNIT 

SHEAR PROCESSES. 

Illustrating possible causes for deviation of 
rate of flow from a hyperbolic sine law at large 
stresses. 

some activation energy in order for flow to 
occur. Concerning the nature of the unit of 
flow and of the energy barrier restricting flow, 
there are no serious limitations, and the theory 
is quite general: the unit of flow may be a single 
molecule or a group of many molecules, and the 
barrier may arise directly from the repulsions 

in which, the molecule is the unit of flow, since 
Al for molecules is of the order of io" 22 c.c. 
(corresponding to a mol volume of 60 c.c.)> a 
good test for the hyperbolic sine law would be 
obtained for stresses such that/ X io~ 2 */2kT is 
of the order of unity. At T sooK. (room 
temperature) this .corresponds to shear stresses 
of the order of 1000 kg. per sq. cm. Clearly, 
such materials as water cannot be studied under 
such high shear stresses. Glasses, however, 
might be subjected to such stresses, and so 
make possible an experimental test of the 
hyperbolic sine law of shear rates. 



between a few molecules or from some more 
complicated mechanism. The usefulness of the 
theory lies in the fact that by fitting it to 
experimental data it is possible to get some 
idea of the magnitudes of AH*, AS* and Al, 
while from these in turn something may be 
learned concerning the fundamental mecha 
nism of the flow. 

One of the difficulties that might be en 
countered in the application of the theory 
entails the measurement of the true value of 
the shearing stress /, which acts on the unit of 
flow. Is this stress the same as <r, the macro- 
scop ically measured shearing stress? For the 
sake of generality, we may say that 

qa- = / [20] 

where q is a stress-concentration factor similar 
to that required in calculating the breaking 
strengths of crystals. 

In the derivation of Eq. 15, it was stated 
that the activation energy for jumping in one 
direction is raised by an amount Alf, while for 
the opposite direction it is lowered by the same 
amount. This is not strictly true, however, 
since the addition of a linear potential field due 
to the force shifts the minima and maximum 
of the resultant field in such a way that the dis 
tances through which the force must act in 
going from the two normal states to the acti 
vated state are not the same. This is made 
clear by Fig. 50. The effect will be most pro 
nounced when the top of the potential barrier 
has a small curvature (Fig. 56) , and will show 
up experimentally in the impossibility of find 
ing a constant, a, that will give a linear plot 
of the creep rate against sinh w. In general, 
however, this effect can be said to be unimpor 
tant, since the shifts that do occur should be of 
the same order of magnitude as the elastic 
strain in the specimen under the given stress, 
and elastic strains are usually small (of the 
order of tenths of a per cent for the macro 
scopic stresses met with here). And if the effect 
is important, but the stresses are so large that 
the range where sinh a<r = e aff /2 is entered, the 
variation in I with the stress can be determined, 
since it is still true that 

/dlnjrate)\ _ qAl r , 

V do- ) T ~ kT L J 

If the situation of Fig. 5& holds, the hyper 
bolic sine law will be drastically wrong, but if I 



WALTER KAUZMANN 



is not very dependent on the stress in the range 
used, Eqs. 21 and 19 will tend to be more valid 
than Eq. 15, even down to relatively low 
stresses; that is, the logarithmic creep law will 
be more nearly correct than the hyperbolic 
sine law even in the range where qAlv is small 
compared with kT. 

Next comes the possibility that the shear 
stress can also act in some ways other than that 
proposed by the present theory in influencing 
the rate of shear. Thus Kanter 8 has postulated 
some effect of the stress on the entropy of 
activation AS*, and Becker 4 has suggested 
that ordinary viscous flow (rate of flow propor 
tional to the shear stress) is caused by a propor 
tionality between X, the distance jumped by a 
unit of flow, and the shear force. Neither of 
these effects appears to have much physical 
significance. Thus it is difficult to conceive any 
mechanism whereby a stress can have as 
marked an effect on the entropy of a regular 
lattice as that required here; and as to Becker s 
suggestion, the explanation of ordinary viscous 
flow afforded by Eyring s formulation appears 
to be far more likely, since the distance jumped 
by a unit of flow* should be related to a lattice 
distance, and lattice distances are known to be 
very insensitive to applied stresses. 

A test of the validity of the assumption that 
the stress affects the shear rate chiefly through 
the addition of a linear, unsymmetrical poten 
tial field to an otherwise symmetrical potential 
energy barrier might be found in a comparison 
of the creep rates for specimens in tensile, 
torsional, and compressional stresses. In ten 
sion and compression, stresses would act normal 
to the shear plane as well as in it, while in tor 
sion there need be no such stresses. If the shear 
rate were the same in all, the effect of stresses 
in other ways than we have pictured would be 
shown to be small. 

Another test might be in the measurement of 
creep rates in tensile tests on single crystals 
that have their most important glide planes 
oriented in different directions with respect to 
the direction of the applied tensile stress. If, 
for a given shearing stress resolved along the 
glide direction in the glide plane, the shear rate 
in that direction is found to be independent of 
the orientation of the plane with respect to the 
tensile stress, our contention will have been 
proved. Rough measurements indicate that 
this is true. 20 



In the derivation of Eqs. 15 to 1 8 it has been 
assumed that shear occurs only by the opera 
tion of a single molecular mechanism. Clearly, 
this need not be true, and in a given specimen 
many widely different mechanisms (such as slip 
along different slip planes, and flow by the 
self-diffusion of single atoms) will always be 
operating simultaneously, each having its own 
characteristic value of Z, X, L, AS*, AH*, A, and 
q. Thus, in general, 



sinh 



17 T 



where the sum is over all of the possible 
mechanisms. Usually only one of the mecha 
nisms (i.e., only one of the terms in Eq. 22) will 
account for the greater part of the observed 
rate. It is entirely possible, though, that a 
single term will not give a major part of the 
shear rate under all conditions: under small 
stresses, for example, one mechanism may 
account for most of the creep, while under 
large stresses another mechanism may be more 
important. Each mechanism, however, con 
tributes something to the total observed shear 
rate under all conditions. A very interesting 
example of this operation of more than one 
shear mechanism has been found by Chalmers 
in the creep of single crystals of tin, and will be 
discussed later. 

Allied to this generalization of the simul 
taneous operation of many different shear 
mechanisms is the question of the interpreta 
tion of strain-hardening and crystal regenera 
tion (Erholung) in terms of the present theory. 
These phenomena, which with the shear rate 
determine the shape of the stress-strain dia 
gram usually used in describing crystal plas 
ticity, are the result of the dependence of the 
constants in Eq. 22 on time and on total strain. 
Thus hardening would result from an increase 
in AH* or from a decrease in AS*, q t A, or I, 
while regeneration would result from changes 
in the opposite direction. Indeed, a study of 
the dependence of these constants on time and 
on strain would probably furnish a valuable 
clue to the nature of the molecular processes 
occurring during strain-hardening and erholung. 

It might be objected that the present theory 
overlooks the possibility that the observed 
creep rates of metals may be due to an equi- 



66 



FLOW OF SOLID METALS AND CHEMICAL-RATE THEORY 



librium between the rate of erholung or anneal 
ing of the metal and its rate of strain-harden 
ing. Such a mechanism of creep has indeed often 
been suggested. A little consideration will show, 



DIRECTION OF SHEAR 

FIG. 6. ILLUSTRATING REQUIREMENT THAT 
SHEAR OCCUR ACROSS AN ENTIRE CROSS SECTION 

OF A BODY IN ORDER THAT STRAINS LARGER 
THAN ELASTIC STRAINS CAN BE OBSERVED. 

however, that this theory can really be in 
cluded in our general treatment. Suppose that 
only structures of a type B in a metal can give 
rise to shear, but that when such a "B struc 
ture" undergoes shear it is converted into an 
"A structure," which cannot undergo shear, 
so that hardening results. But A structures can 
spontaneously go over into new B structures by 
the process of erkolung. Then if the rate of 
formation of A structures from B structures is 
large compared with the reverse process, the 
rate of shear will be determined by the rate of 
the erholung. We have supposed that the 
erholung process by itself does not result in 
shear; its rate cannot then be affected by the 
action of shearing stresses, and the result 
should be a creep rate independent of the stress. 
These are commonly not observed, although 
one possible example may appear later on. If, 
on the other hand, the erholung did result in 
shear, its rate could be treated by means of 
Eqs. 15 or 18. 

Before examining available data in the 
light of Eqs. 15 and 18, it might be well to 
mention one feature of shear that must be 
kept in mind, particularly when studying 
crystalline solids: When a plane along 
which shear occurs has an area A, and when 
shear amounting to 6 occurs over only a 
small portion of this plane having an area 
a, the macroscopically observed shear over 



the entire area A will not be given by 5 t 

The release of shear stress in a part of a 
shear plane can only give rise to an ob 
servable shear corresponding to an elastic 
strain over the remainder of the plane. To 
illustrate this, suppose that in the bar pic 
tured in Fig. 6a a shearing motion amount 
ing to 5o cm. has occurred over half the area 
of each of n planes making an angle of QO 
with the axis of the bar. Suppose that by 
virtue of this motion all of the shearing 
stress has been released over these regions. 
Then the lateral displacement of the ends 
of the bar in the direction of the shear stress 
will not be given by n5o/2, but instead will 
be given by the elastic shear strain of a bar 
from which these portions of released stress 
have been cut out (Fig. 6b). Since elastic 
strains are always small, large plastic shear 
(from, say, a small percentage up to the 
large percentages that may be observed 
in single crystals) can only result from a 
shear all of the way across the cross section 
of the bar. 

The significance of this fact is that if 
potential units of flow are not suitably 
distributed through a shear plane, their 
movement cannot result in the expected 
macroscopically observable shear. In liq 
uids and even in amorphous solids, this 
requirement is unimportant, since it is 
readily fulfilled by the mobility of the 
molecules and the homogeneity of the 
material. Crystalline solids, however, prob 
ably are not homogeneous, and their con 
stituent atoms and molecules are not very 
mobile. Potential units of flow probably are 
scattered relatively sparsely throughout 
the specimen. Therefore, this requirement 
is likely to play an important role in deter 
mining the rate of now of solids. 

APPLICATION OF GENERAL 

THEORY TO AVAILABLE DATA 

ON CREEP OF METALS 

In the following pages an attempt will 
be made to interpret some of the available 



WALTER KAUZMANN 



6 7 



experimental data on the flow of crystalline 
metals in the light of the general equations. 
The rate of shear of a crystalline metal is 
measured usually by determining the rate 
at which a bar under tensile stress increases 
in length; the time rate of increase of ten 
sile strain (i.e., the creep rate), u = ~r 

d-i 

(where e is the elongation per unit length) , 
is related to the shear, 7 max , occurring along 
the plane of maximum shear stress by 21 



t Tt COS <j>i COS tyi 



[2 7 ] 



" - 3 dt L ^ J 

while the shearing stress, 0- max , in this 
plane in terms of the tensile stress r is 

<r max = r/2 [24] 

Substituting Eqs. 23 and 24 in Eq. 18, the 
relationship between tensile stress and 
creep rate becomes: 



__ 

3L h 



[ 2 cj] 



Creep rates have also been measured with 
tests in torsion. 31 In this case the torsional 
creep rate is given directly by Eq. 18, 
where a is the applied torsional stress. 

In single crystals, where shear can occur 
in various glide planes that make angles 
4>i, 02, . . - with the direction of the 
tensile stress, and along different glide 
directions in these planes which make 
angles ^i, ^2, with the projection of 
the direction of the tensile stress in the 
glide plane, the tensile creep rate will be 
the resultant of several simultaneous 
mechanisms. Remembering that the re 
solved shear stress en for one of these glide 
systems is related to the tensile stress T by 



- sin 

2 



cos 



[26] 



and that the amount of shear 7; in such a 
system results in a contribution to the 
tensile creep of amount 



we see that the total tensile creep of a 
single crystal is given by 

, \ikT 

COS <pi COS W J 7- < 



all glide 
systems, * 



where /,-, -L, AF**, etc., are the quantities 
typifying each glide system. If tests are 
carried out on any one single crystal, pre 
sumably most of the creep will usually 
result from the slip occurring in one glide 
system, and only one term of Eq. 28 will 
be required to describe the creep behavior 
of the particular crystal. This may not, 
however, be true at all temperatures and 
at all stresses for a given crystal, and in 
crystals with different orientations with 
respect to the tensile stress, different glide 
systems will operate in general. 

In order to evaluate the constants appearing 
in Eq. 18, the data employed must include 
measurements of creep rates for a range of 
stresses at each of several temperatures. The 
procedure that has been used here is illustrated 
by Fig. 7. using the data of Chalmers 22 and 
Tyte 26 on tin. The logarithm of the steady- 
state creep rate at a given temperature is plot 
ted against the stress, and from the slope of 
this plot, which usually is a straight line, can be 
derived a value of qAL By extrapolating the 
straight lines back to zero stress, the values of 
the logarithms of 



are then found for each temperature.* These in 
turn are plotted against i/T to give values of 
AH* (from the slope of the resulting straight 



* The validity of this extrapolation has 
nothing at all to do with the question of 
whether there is a finite creep rate at zero 
stress; the extrapolation is merely a con 
venient mathematical method for evaluating 



the quantity ~ ^*/*-A**/i" at the tem- 
perature in question. 



FLOW OF SOLID METALS AND CHEMICAL-RATE THEORY 



line) and of 



(from the intercept of the straight line at 



cannot be further resolved to give individual 
values of q, A, I, X, L and AS*. But even this 
incomplete evaluation of the constants is very 
useful in understanding the creep mechanism 
in metals. 



O-POLYCRYSTALLINE flN(23) 
=SINGLE TIN CRYSTAL(2Q) 




SLOPE = 33.4 X 10 3 qAl=6l300 Af 
INTERCEPT = 1.4 8 x 10 B 



L _- MA=478 
INTERCEPT =I.6*IO~*| 

SLOPE=555 * 10" <jAl= 16300 A a ; 
INTERCEPTS! *IO~* 

SLOPE =520*l(r 3 qAl= I4500A 03 ; 
INTERCEPT =4.5*1" 



INTERCEPT=3.2* ltf e 
3 ; INTERCEPT = 2.8* IO 6 
= 51 QOA 3 ; INTERCEPT- I.I l(5 a 
3 , INTERCEPT = .3 1 I0" a 



200 



400 600 800 1000 1200 



1400 



TENSILE STRESs(! m /nun*) - 

FIG. 7. CREEP BEHAVIOR or TIN. 22 - 26 METHOD OF EVALUATING qAl PROM OBSERVED DATA, 



i/T = o, the variation of the factor kT/h with 
temperature being negligible compared with 
that of the factor e ~* a * /RT in the same range, 
so that kT/h may be regarded as a constant, 
with T the mean temperature in the range). 
This is, of course, as far as we can go in deter 
mining the constants from the experimental 
data; the factors qAl and 



The greatest difficulty in this procedure is 
to obtain data sufficiently accurate to give 
reasonably smooth curves when the logarithm 
of the rate is plotted against the stress, and 
more especially, when the logarithm of the 
rate intercept at zero stress is plotted against 
i/T. The latter plot, particularly, is likely to 

be erratic, so that the values of AH* and 7 c AS * /B 

Ju 

obtained are not as reliable as the values of 
qAl, although their orders of magnitude are 



WALTEE KAUZMANN 



6 9 



certainly correct in most cases, and this is 
what is most important at present. 

Concerning the factor X/L, it would be 
expected that ordinarily this should be of the 
order of magnitude of unity. Any marked 



Chalmers data on the creep of single crys 
tals of tin, it has been assumed that <t> = 
45 and ^ = o, as this was indicated by 
him to be nearly true. Owing to the large 
number of equivalent glide systems in tin, 



TABLE i.Data on Different Materials 



Number 




Reference 


A#*, 


AS*, 


AF*. Cal. per Mol, at 


of Curve 
Giving qAl 
in Fig. 9 


Substance 


No. 
(Page 81) 


Cal. per 

Mol 


Cal. per 

D !X r 




ioooK. 


300K. 


la 


Tin, single crystal 


22 & 








37,5oo 


i 


Lead 


23 


3,800 


-99 


103,000 


34,000 


2 


Lead 


24 


2,3008,400 


-93 to -83 


95,000-91,000 


30,000-33,000 


2d 

3 


Lead 
Tin, poly crystalline 


25 
26 


13,000 
5,350 


-74 
-80 


87,000 
85,000 


35,ooo 
29,350 


4 
5 


Zinc, High- Grade 
Zinc, Brass Special Grade 


27 

27 


31,400 
16,500 


i 
49 


32,000 
65,000 


31,000 
31,000 


6 


60-40 brass 


28 


15,300 


-76 


91,000 


38,000 


7 


0.4 % C steel 


29, p. 244 


28,500 


-75 


103,000 


51,000 


8 


0.5 % Mo steel (0.5 % Mn, 


29, p. 72 


45,000 


-55 


100,000 


62,000 




0.25 % Si) 












9 

TO 


0.34 % C steel 
0.5 % Mo steel (0.2 % Mn, 


30 

29, p. 78 


26,300-29,400 
17,000-21,000 


60 to 52 
90 to 86 


86,000-81,000 
107,000 


44,000-45,000 
44,000-48,000 




1.4 % Si) 












11 


1 8 % Cr, 10 % Ni steel 


29, p. 216 


5,000-15,000 


97to 108 


112,000 


37,00046,000 


12 


Cast steel 


31 


102,000 


+ 14 


88,000 


98,000 


13 


C steel 


a 


60,100 


-42 


102,000 


73,ooo 


14 


C-Mo steel 


a 


41,500 


-60 


102,000 


59,000 



Data obtained from Mr. N. L. Mochel, of the South Philadelphia Works of the Westinghouse Electric 
and Manufacturing Co., and from Mr. P. G. McVetty and Mr. D. W. Gunther, of the Westinghouse Research 
Laboratories. The author wishes to express his thanks for the opportunity to use these data. 

& The data used here were given as a plot of creep rate against stress in Fig. 3 of this reference, but the scale 
of stresses in this plot is obviously in error. Prom the context of the article, however, it is probable that the zero 
point and the point marked 100 gm./mm. 2 are correct, so in using these data, the rest of the scale was taken to 
conform to these two points. 



deviation from this, as from a wide spacing of 
the glide planes (LX), can best be re 
garded as an entropy effect, reflecting, say, a 
sparse concentration of the units of flow. 
Therefore, we can regard it as true that L = X 

and write AS* = R lo ge Q e *** *), where AS* 
now includes a possible concentration effect. 

The values of A#* and AS*, and values 
of AT?* at 3ooK. and ioooK. derived from 
these, for a number of different materials 
for which creep data are available, are 
given in Table i. The values of qAl were 
found to be strongly dependent on the 
temperature in most cases, so these are 
plotted semilogarithmically against the 
temperature in Fig. 8. For all of the poly- 
crystalline materials represented here, the 
creep rates used were the so-called " second- 
stage" creep rates, but for the tin single 
crystals initial creep rates were employed. 

In the values of qAl and AF*, from 



one at least will find itself nearly so oriented, 
so will cause most of the observed creep. 

INTERPRETATIONS OF THE DATA 

The data summarized in the preceding 
paragraphs nearly all show the same gen 
eral characteristics. Taking these char 
acteristics as typical of metallic crystalline 
flow in general, we shall now try to under 
stand their significance in terms of a 
general molecular picture. An attempt will 
also be made at interpreting some of the 
less general features associated with 
metallic now, such as the reason for the 
much greater softness of lead than of iron, 
and for the great softness of single crystals; 
the nature of the microcreep found by 
Chalmers in single tin crystals; the possible 
effects of grain boundaries and impurities; 
the role of self-diffusion in creep; and creep 
at very low temperatures. It must be 
understood, however, that some of the 



FLOW OP SOLID METALS AND CHEMICAL-RATE THEORY 



interpretations advanced are somewhat 
speculative, particularly with respect to the 
less general characteristics mentioned 
above, and that sometimes they are made 



value of qAl found for all of the metals. 
Thus, if q = i (no stress concentration), 
and if A is the cross-sectional area of a 
single metal atom, and if I is one lattice 



o 

3 

u. 
O 



I0 8 



LJ 



200 160 



120 



-TC 
80 



60 40 



R = 1, 170. AH* 5350 CALS. 

5^ = 3.71 XKT 5 

AS*=-79.6 CALS/ DEC 




LO] 



2.0 



2.4 



2.8 



3.2 



3.6 



FIG. 8. CREEP BEHAVIOR OF TIN. 22 * 26 METHOD OF EVALUATING AH* AND AS* FROM OBSERVED 

DATA. 



on the basis of rather meager shreds of 
experimental evidence. But the purpose of 
the present paper is predominantly to 
illustrate the nature of the information 
derivable through the application of 
Eqs. 15 and 18 to suitable experimental 
data rather than to develop any concepts 
that must be regarded as very definite 
with regard to metallic flow. So perhaps 
this is excusable, especially in the light 
of the fact that the speculations are capable 
of rather direct verification or disproof, as 
the case may be, through suitably designed 
experiments. 

One of the most striking characteristics 
of the experimental data is the very large 



distance in a crystalline metal, values of 
qAl of something like 10 A. 3 would gener 
ally be found for metals (i A. = io" 8 cm.). 
Instead, values 100 times or more as large 
are usually found. It is this large value of 
qAl, together with the large stresses required 
to give rise to detectable rates of flow, that 
causes the creep of metals to depend upon 
the stress exponentially rather than linearly. 
The second striking feature in the ex 
perimental data is the generally very small 

value of y e^ s */ R . Ordinarily it would be 

JU 

expected that if L and X corresponded to 
lattice distances, AS* would be close to 
zero, and certainly no larger than io. 



WALTER KAUZMANN 



Instead large negative values are found in 
the neighborhood of 50 to 100 cal. per 
degree per mol in 12 cases out of 14, corre 
sponding to values of - e A5 */-K a b out 

jL 

io- 10 to io- 20 . 

These two facts could perhaps be ex 
plained if the unit shear processes occur 
only at points of great stress concentration 
(q of the order of several hundred), and if 
this unit process were a simple passage of 
metal atoms by one another as in self- 
diffusion. Such an interpretation is sup 
ported by the fact that AH* is found to be 
usually of the same order of magnitude as 
the activation energies of diffusion proc 
esses in solid metals. The very small value 

of - e AS */ R might then be a consequence of 

Ju 

the small number of points at which the 
value of q is as large as it is found to be here, 
since for the reasons discussed on page 66, 
in order for flow to occur at all in a given 
plane, units of flow must be properly 
distributed all over the plane, and owing to 
their scarcity here, this is very improbable 
of occurrence, so that e AS * /R is very small. 

This last argument, however, is not at 
all convincing, although it seems necessary 
if the explanation for the large value of qAl 
is taken to be associated with a large value 
of q and if the observed values of AS* are 
to be accounted for. The proper distribu 
tion of sites of unit shear processes requires 
that nearly every lattice site in a given 
shear plane at some time be the location 
of a unit shear process. This means that 
the points of stress concentration must 
readily wander over the given shear plane, 
since a concentration on the order of one- 
hundredfold is utterly impossible at all 
points simultaneously. This is in turn 
highly unlikely, since such concentrations 
of stress as do occur are probably pro 
duced by cracks and other similar im 
mobile features. 

It might also be suggested that in poly- 
crystalline material the rate of shear is 



governed by the shear of those crystals that 
are least favorably oriented with respect to 
the tensile stress. The surrounding crystals 
will then be continually relaxed and the 
stresses would concentrate on the few 
unfavorably located crystals. Evidently 
something like one crystal in one hundred 
must be so oriented in any given cross 
section for this explanation to apply. The 
difficulty here is twofold: first, the even 
larger qAl value of single crystals of tin 
must be explained, and second, unless the 
unfavorably oriented crystals form some 
kind of interlocking network among them 
selves (which is highly improbable), they 
will affect the creep rate of the specimen in 
the same way as the addition of one per 
cent or less of sand to water affects the 
viscosity of the water (an effect that is 
very slight) . 

Finally, it is not readily understood why 
stress concentrations should depend so 
strongly on the temperature as the be 
havior of qAl would indicate. 

The conclusion thus seems inescapable 
that the large value of qAl cannot arise 
from a large value of q t and it will in fact 
probably be safe to take q as equal to 
unity.* That I could be large enough to 
account for the values of qAl is even less 
admissible, for it is inconceivable that an 
atom could in one leap move something 
like 100 lattice distances before arriving at 
the activated state. 

There remains, therefore, only the pos 
sibility that the area A is much larger than 
the cross section of a single atom. This 
explanation is very attractive, since it is 
what we should expect if the unit shear 
process occurred through the formation of 



* It does, however, seem very possible that 
large q values may be of some importance in 
determining the now of polycrystalline ma 
terials immediately after loading, when there 
is probably a great stress concentration at a 
relatively few points in the material. These, 
however, are soon evened out by the plastic 
now. It is interesting that correspondingly 
rapid initial creep rates do not seem to be 
observed in single crystals. 



PLOW OF SOLID METALS AND CHEMICAL-RATE THEORY 



" dislocations " such as are illustrated 
schematically in cross section in Fig. 10 
and have been employed previously by 
Taylor, Orowan, Polanyi and others to 



I0 5 



Then if the lattice distance is X, the area 
of the top of the block that undergoes shear 
due to the motion of the dislocation is n\d 
and the actual shear stress acting on the 



3- LOW STRESSES 



3,-HIGH STRESSES 



SPACING OF 
OPAQUE SURFACES 
IN ALUMINUM 
SINGLE CRYSTALSCTAYLOR)^ 
(RIGHT HAND SCALE) 



IGH STRESSES 
2-LOW STRESSES 




200 



4OO 



600 



800 



T C 



FIG. 9. VALUES OF qAl FOR DIFFERENT METALS AT DIFFERENT TEMPERATURES. 
The key to the numbers is in Table i. All the curves except the one marked by solid triangles 
correspond to the left-hand ordinate scale. 



explain the plasticity of crystals. It must 
be emphasized that the regions of disloca 
tion in Fig. 10 extend over a considerable 
distance in the direction perpendicular to 
the plane of the paper. Let us call this 
distance d. Suppose the block of material 
through which the dislocation travels in 
the plane of the paper is n atoms wide. 



block is (assuming q i) n\d<r. If at any 
instant the distance in the shear direction 
moved by a point on the upper surface 
of the block relative to that of a point 
on the lower surface of the block is fj,, it 
can be said that /-c = Xa, where a is the 
fraction of the distance through the block 
moved by the dislocation. A plot of the 



WALTER KAUZMANN 



73 



energy against /z would then look like 
Fig. n, and the flat part of the curve 
between /z = o and /z = X is very long 
compared with the rest of the curve when 
n is large; this corresponds to the disloca 
tion when located inside the block. The 
energy basins, on the other hand, corre 
spond to the dislocation when it is close 
to or at the surface of the block, and this 
in turn corresponds to values of /z of ap 
proximately \/n\ that is, the distance 
through which the two parts of the block 
must be moved relative to one another in 
order to introduce a dislocation into the 
block is something like \/n. The work done 
by the shear force in lowering the activation 
energy is thus approximately 

and this must be equal to qAlo-. Since X 2 
is of the order of io A. 2 and q is close to 
unity, and qAl usually has value of ioooA. 3 
or more, d must be of the order of magni 
tude of 100 A. or more. 

It is to be noted from Fig. 9 that the 
logarithmic rate of change of qAl with 
temperature is strikingly similar for nearly 
all of the metals and this even when we 
have to deal with metals as widely different 
as iron and lead. This dependence of qAl 
on the temperature is best ascribed to a 
marked increase in d with increasing 
temperature; at higher temperatures much 
wider blocks are evidently involved in slip. 
This might be compared roughly to the 
effect of temperature on the separation of 
the surfaces that prevent the transmission 
of dislocations ("opacities") postulated 
by Taylor to account for strain-hardening, 
and would presumably arise in much the 
same way. It is interesting that if gX 2 for 
single crystals of tin is taken to be io A. 2 , d 
is found to be of the order of magnitude of 
io 4 A., which is of the same order of magni 
tude as the separation of the surfaces of 
misfit found by Taylor for other single 
crystals. In Fig. 9 there has been included 
a plot of the temperature dependence of 



this separation of the surfaces as found 
by Taylor for aluminum. The logarithmic 
rate of change of this separation with tem 
perature is seen to be about the same as 




POSITIVE DISLOCATION 




d e f 

NEGATIVE DISLO CATION 

FIG. io. SHEAR OF CRYSTAL BY MEANS OF 

MOTION OF DISLOCATIONS. 

that for qAl for most of the other metals. 
The conclusion might be that the same sort 
of barrier that stops the forward movement 
of dislocations also acts to prevent their 
extension in the direction perpendicular to 
their motion. 

We come now to a consideration of the 
observed values of AH* and AS*. It seems 
fairly obvious that the energy required to 
form a dislocation of length d should, at 
least in part, be proportional to the length 
of the dislocation. That is, 



AH* = AH\ + AH*id 



[i] 



where AH*o might be an energy associated 
with the strain occurring at the ends of the 
dislocation, and AH* id is the part of the 
energy depending on the length of the dis 
location. As d is markedly dependent on 
the temperature, this should give rise to 
some dependence of the activation energy 
for creep on the temperature. To a rough 
first approximation, it is true for most of the 
metals considered here that qAl, and hence 
d, changes linearly with the temperature 
over the range of temperatures usually 
used. That is, 



74 



FLOW OF SOLID METALS AND CHEMICAL-RATE THEORY 



[ii] small (i.e., making for an apparently 
negative entropy of activation in Table i) 
might be the following: It is very probable 
that dislocations occur most readily at 
only a very few places in an actual crystal 
(say at those positions at which AH* is 
rjyi particularly low because of some lattice 
imperfection). Furthermore, because of the 
That is, in addition to the constant factor condition discussed earlier in this paper 



d = do + d T 

Substituting Eqs. i and ii in the creep 
equation, No. 18, we find 



5 = _ e *s*/R[ e -&H*idi/R] e 
L h 



RT 



I 

UJUJ 

fez 





FIG. ii. POTENTIAL-ENERGY CURVE FOR A DISLOCATION. 



involving the entropy, there is a constant 
factor (shown in brackets in Eq. iii) in 
volving the rate of change of activation 
energy with temperature. This factor will 
show up as an apparent negative entropy 
of activation.* Furthermore, the "activa 
tion energy," when determined as in this 
work, appears to be only Afi r * + doAH*i, 
which is presumably much less than the 
true energy of activation. The temperature 
variation of the activation energy can 
thus be made, qualitatively at least, to 
account for both the very small activation 
energies and the negative "entropies of 
activation" that have been observed. A 
more quantitative analysis of the activa 
tion energy will be made at a later date, 
using equation i directly with values of d 
as estimated from observed values of qAL 

A further factor tending to make j e^ s * R 

* A similar apparent entropy effect is found in 
the study of the conductivity of ionic crystals, 
although here an increase in the temperature 
results in a decrease in activation energy, so 
gives rise to a positive "entropy of activation." 
See for example, F. Seitz: Modern Theory of 
Solids, 555. New York, 1940. 



(p. 66), only when these imperfections are 
suitably distributed across a potential 
shear plane in the crystal can macro- 
scopically observable shear occur in that 
plane. This means that only relatively few 
of the planes in the crystal can function as 
shear planes, so that L is much greater than 
X; this is probably the cause of distinct 
slip bands in a stressed crystal. If these 
bands are spaced by, say, io~ 4 cm., and 
if X is of the order of io~ 8 cm., the factor 
\/L would have a value of around io~ 4 , 
giving a contribution to what is given as 
entropy of activation in Table i of some 
thing like 15 to 20 cal. per deg. 
per mol.* 



* The results of D. Hanson and M. A. 
Wheeler 35 tend to indicate, however, that when 
a metal is allowed to creep slowly (under small 
stress) to a given extension (i.e.. in long-time 
creep tests such as those from which the data 
used here were all obtained) slip bands are not 
observed even though they may form when a 
similar specimen is stretched rapidly to the 
same extension. Presumably, therefore, the 
bands are not so widely spaced under creep 
conditions, and this factor may be less impor 
tant here. 



WALTER KAUZMANN 



75 



The general features of metallic flow thus 
seem to indicate that the flow occurs by 
means of dislocations which form at only 
relatively few points in the metal lattice, 
and whose lateral dimension depends 
strongly on the temperature, in a manner 
probably similar to that of Taylor s 
"opaque surfaces." We now turn to some 
more specific questions related to metallic 
flow. 

First, why does solid lead, for instance, 
flow so much more easily than steel? 
Table i shows that for iron and steel at 
ioooK. the values of AF*, the free energies 
of activation for the unit shear processes, 
would be nearly the same for the two (as 
suming lead to remain solid at this tem 
perature). Indeed, the values of AF* are 
surprisingly similar for all metals at 
ioooK. Therefore, the origin of the differ 
ence cannot lie here. Instead, we see that 
lead has a very much larger value of qAl 
than steel (Fig. 9) perhaps 100 times as 
large at any given temperature. This, if 
what has been said above is true, is prob 
ably due to a much greater lateral extension 
of a region of dislocation in lead than in 
iron, which is, in turn due to the presence 
of fewer or less effective opaque surfaces. 
In fact, it seems to be true that the greater 
part (but not all) of the difference in the 
creep rates of solid metals at ioooK. is 
due to this very wide difference in the 
values of qAl, hence of the lateral extension 
of dislocations, which they show at that 
temperature. 

It is a well-known fact in metallurgy 
that soluble impurities cause a metal to 
harden much more than do insoluble, inter- 
granular impurities. An explanation of this, 
perhaps, is to be found in the following: 
when a dislocation, on extending itself in 
a lateral direction (direction perpendicular 
to the plane of the paper in Fig. io)> meets 
a foreign atom, its progress in that direction 
is arrested. This tends to decrease A and 
perhaps also to increase AH*, which in turn 
makes the material flow much more slowly 



and thus become harder. An impurity at a 
grain boundary will be able to have no such 
effect. (The validity of this suggestion can 
be readily tested by means of appropriate 
experiments, of course.) 

The available data on the creep rates of 
single crystals are inadequate for the com 
plete application of Eq. 18. The measure 
ments by Chalmers on tin crystals at room 
temperature, however, make it possible 
to find values for qAl and AF*. These values 
can then be compared with the data of Tyte 
on polycrystalline tin (the effects of differ 
ences in purity being assumed, perhaps 
wrongly, to be small). It is found that qAl 
is about 10 times as large for single crystals 
of tin as for the polycrystalline tin at room 
temperature. On the other hand, AF* is 
also somewhat larger for the single crystals. 

The question is often asked: Would a per 
fect crystal be extremely soft or extremely 
hard? Is the ease of flow of single crystals 
a property that depends on structure im 
perfections, or is it a property of the perfect 
lattice? The answer to this question may be 
indicated in the facts mentioned above. 
The larger value of qAl in single crystals 
probably is due to the possibility that dis 
locations can extend over a much greater 
Jateral distance, and the more perfect is 
the crystal, the greater should this distance 
be, since there will be fewer "opaque sur 
faces" to prevent the extension of disloca 
tions. Yet, if dislocations can form most 
easily at only certain points of imperfection 
(and we have seen that this is likely), the 
number of these points will decrease as the 
crystal becomes more perfect. This will 
cause AS* to become more negative and 
may also tend to increase AH*, either or 
both of which will cause AF* to increase. 
Whether or not a crystal will become more 
plastic as it becomes more perfect will 
then depend upon which of these two 
opposing effects will predominate in the 
process of approaching the perfect crystal. 
And although it would seem that a really 
perfect crystal should not be very plastic, 



7 6 



FLOW OP SOLID METALS AND CHEMICAL-RATE THEORY 



in actual practice qAlcr (where a has the 
value corresponding to the stresses ordi 
narily used) evidently increases somewhat 
more rapidly than AF*, so that it appears, 
from experiments, that the more perfect a 
crystal is, the softer it becomes. 

The temperature coefficient of creep is 
evidently a composite effect, since it arises 
as much from the existence of an activation 
energy for creep as from the great change 
in qAl with temperature. This means that 
if an attempt is made at evaluating the 
activation energy by using creep measure 
ments at a single stress and different 
temperatures, a much larger energy will 
be found than is actually operative. 

CREEP AND SELF-DIFFUSION 

Chalmers, in his study of the creep rate 
of single tin crystals, was able to show that 
at low stresses in these crystals the creep 
rate is directly proportional to the applied 
stress, rather than to an exponential 
involving the stress, whereas at higher 
stresses (above about 200 grams per sq. 
mm.) the rate increases much more rapidly 
with the stress. Chalmers assumed that 
two different mechanisms were acting: he 
called the creep whose rate is proportional 
to the stress "micro-creep" and that which 
increased more rapidly "macro-creep." 
The question immediately arises, whether 
the two types of creep really involve 
different mechanisms or are merely the 
consequence of the hyperbolic sine law 
of Eq. 15. An attempt to fit Chalmers 7 data 
into a hyperbolic sine equation leaves no 
doubt of the correctness of his interpreta 
tion that there are two distinct mechanisms; 
we are clearly dealing with a case in which 
the total rate must be given by a sum of 
two separate rates, as in Eq. 22. 

From the fact that the creep rate is pro 
portional to the applied stress, and knowing 
the range of stresses for which this is true, 
we can place an upper limit on the value 
of qAl which characterizes this type of 
creep. (The problem is simply to find the 



largest possible value of a that in a given 
range of values of x will enable sinh ax to 
be very nearly proportional to x.) It turns 
out that qAl must be no greater than 
5000 A. 3 , or about one-tenth of its known 
value of 63,600 A. 3 for macrocreep. This is 
still a large value, but is it not possible 
that its actual value is something closer 
to the much smaller one (~ 15 A. 3 ), which 
would correspond to single atoms acting as 
units of flow?* That is, it is quite possible 
that microcreep occurs simply by the same 
process as self-diffusion, and that this is 
the reason for the observed proportionality 
between creep rate and stress. 

In order to investigate this problem fur 
ther, it is necessary to look into the rela 
tionship that might be expscted to exist 
between diffusion rates and shear rates. 
Eyring 3 has shown that, given the rate at 
which a substance diffuses through itself 
(the rate of self-diffusion) and if the unit 
molecular process in shear is the same as 
that in diffusion, it should be possible to 
calculate a corresponding shear rate or 
viscosity coefficient. This is readily under 
stood as follows: 

The net rate at which molecules move 
in a given " forward" direction (the dif 
fusion rate) is given by the difference 
between the total number of molecules 
that jump in the forward direction in one 
second, and the number that jump in the 
opposite direction in the same time. Since 
the number of molecules that jump through 
a given plane is proportional to the total 
number of molecules present that are 

* A further indication that microcreep does 
not occur by the movement of dislocations is 
found in the remarks (p. 65) to the effect that 
potential energy curves such as those typical 
of the movement of dislocations (Fig. n) tend 
to give creep rates that depend exponentially 
on the stress at much lower stresses than would 
be expected from the hyperbolic sine law. Since 
flow according to this law is required in order 
that creep rates can be proportional to shear 
stresses, it must be suspected that a potential 
energy curve such as that in Fig. n, and hence 
flow by means of dislocations, does not occur 
in the microcreep of tin. 



WALTER KAUZMANN 



77 



capable of jumping through this plane, the 
number jumping forward through a plane 
having a unit area per second is 

n+ = koc\ [30] 

where ko is the rate at which a single mole 
cule jumps (number of jumps per second 
by one molecule), c is the concentration of 
the molecules behind the plane, and X is the 
distance moved by a molecule in one jump 
(j&o is the same as n in Eq. 10). c\ thus rep 
resents the total number of molecules 
behind the plane that are capable of jump 
ing through it in one leap. Similarly, the 
number of molecules jumping in the op 
posite direction is 



[31] 



dc . 



where y- is the gradient of the concentra 
tion in the forward direction, so that 
( c + j- X ) is the average concentration 

of molecules in front of the plane. Then 
the diffusion rate is given by 

and the ordinary diffusion coefficient A is 
A = X 2 [33] 

On the other hand, it has been shown that 
the shear rate may be given by 



s ~ TT 

or, substituting Eq. 33, 



s = 



XL 



[34] 



[35] 



If Newtonian flow is involved, sinh 
qAlcr/kT can be replaced by qAla/kT and 
the viscosity coefficient calculated as 

a- \LkT 



for diffusion processes in ordinary liquids 
when reasonable values of X, Z, etc., are 
used; it will now be applied to a calculation 
of the rate of flow of solid and molten lead. 

The rate of self-diffusion of lead (thorium 
B) into molten lead at 343C. has been 
found 32 to be 2.2 sq. cm. per day =2.55 
X io~ 5 cm. per sec., and the viscosity of 
lead at 343C. can be found from a short 
extrapolation of a logarithmic plot of 
viscosity data 33 to be 77 = 0.0266 poises. 
Using the observed A and taking q = i and 
X = L = il = -\/A = 3.1 X icr 8 cm., f\ is 
calculated to be 0.107 poises. This is too 
large by a factor of 4 (not too serious an 
amount when the uncertainties in X, L, /, 
and A are taken into consideration). 

Turning to the case of solid lead, let us 
compare the creep rate of 4 X io~ 6 per day 
at a tensile stress of 200 Ib. per sq. in. meas 
ured by McKeown 23 at room temperature 
(3ooK.) with that expected from Eq. 35. 
This corresponds to a shear rate of 3 X io~ 6 
per day and a maximum shear stress of 
100 Ib. per sq. in. The diffusion rate of lead 
into solid lead at 3ooK. can be calculated 
from an extrapolation of the data of Seith 34 
to be A = 3.5 X io~ 15 sq. cm. per day. 
Taking q = i, Al = i5A. 3 , and X = L 
= 3A., sinh qAh/kT 2.5 X io~ 3 and a 
shear rate of 1.9 X io~ 2 per day. This is 
5000 times faster than the rate observed 
by McKeown, whereas a shear rate cal 
culated for molten lead would be 4 times 
too slow. Furthermore, if the creep of lead 
were occurring by means of a self-diffusion 
mechanism under the present conditions 
of stress and temperature, the conditions 
leading to ordinary viscous (Newtonian) 
flow would obtain and the creep rate would 
be proportional to the applied stress, which 
was not found to be true by McKeown.* 

Therefore, the observed creep of solid 
lead does not occur by any simple self- 



This equation has been found to hold well 



* Evidence for the existence of ordinary 
viscous flow in lead at low stresses may, how 
ever, be found in the data of Hanfstengel and 
Hanemann 36 and of Moore and Betty. 37 



7 8 



PLOW OF SOLID METALS AND CHEMICAL-RATE THEORY 



diffusion mechanism while the flow of 
molten lead probably does involve such a 
mechanism; and conversely, either the unit 
molecular self-diffusion process in solid lead 
does not involve shear, or self-diffusion in 
solid lead cannot be the simple, homogene 
ous process postulated above, but occurs 
chiefly along preferential paths in the 
crystal paths so located with respect to 
one another that macroscopically observ 
able shear can result from movements 
along only at most one out of every 5000 
of the paths in the sample of lead studied 
by McKeown. 

If the microcreep discovered by Chal 
mers for single crystals of tin also arose 
from a self-diffusion mechanism proceeding 
chiefly along certain paths in the crystal, - 
we might be able to understand certain 
other characteristics of microcreep found 
by Chalmers. For instance, he found that 
associated with microcreep there is a rapid 
strain-hardening, and the total extension 
obtainable through the action of the micro- 
creep mechanism is more or less independ 
ent of the applied stress. On the other hand, 
the hardening so produced disappears 
within a few hours, or after a larger stress 
is applied for a short time. Chalmers ex 
plained this by saying that the microcreep 
is the result of the flow of dislocations 
already present in the metal, that this 
number is limited, and that new disloca 
tions form very slowly at such low stresses; 
therefore, only the amount of creep corre 
sponding to the number already present 
is possible. In the light of the present 
theory, however, it seems more likely that 
the rapid strain-hardening associated with 
microcreep is caused by the exhaustion of 
the diffusion paths favorably located for 
shear. Since self-diffusion occurs because 
of the presence of empty lattice sites in the 
crystal, this exhaustion might be due to a 
depletion of all of the empty sites along 
the given paths. The regeneration of the 
crystal on standing would then result 
from a slow seeping of new empty sites 



into the paths from the surrounding 
medium, while the regeneration due to the 
application of larger stresses might result 
from a readjustment that makes new paths 
of self-diffusion effective in causing shear. 
Chalmers found that after a crystal had 
undergone all of the microcreep of which 
it is capable there was a small residual creep 
which proceeded exceedingly slowly and 
whose rate seemed to be independent of 
the stress. (This last type of creep was not, 
however, very thoroughly studied, owing 
to the experimental difficulties necessarily 
involved.) Chalmers ascribed this residual 
creep to the slow formation of new disloca 
tions in the crystal. Such an explanation, 
however, is clearly inadmissible according 
to the present theory; the rate of formation 
of dislocations should be markedly affected 
by the applied stresses, because, as we have 
seen, the applied stresses can do a great 
deal of work to aid in the formation of any 
dislocation that will ultimately result in 
shear, and this, by altering the activation 
energy, alters the rate. Instead, we would 
suggest that this residual creep is caused 
by the slow seeping of holes (empty lattice 
sites) into the effective diffusion paths as 
mentioned above. Since the direction of 
this seepage is perpendicular to the shear 
direction, the applied forces cannot help 
the holes appreciably in their motion into 
a path of shear, and the resultant shear 
rate is independent of these stresses insofar 
as the rate of seepage of a hole into a path 
is slow relative to the rate of the diffusion 
of the hole down the path. (Cf. the over-all 
rate of two successive chemical reactions.) 

CONTRAST WITH OTHER THEORIES 

The numerical results of the application 
of Eq. 1 8 to actual creep data may be said 
to tend to verify the theory that metals 
deform by means of dislocations; on the 
other hand, any theory based upon ideas 
of self-diffusion, such as that of Kanter, 8 
does not seem to be valid under most 
conditions. 



WALTER KAUZMANN 



79 



Taylor s theory of strain-hardening 6 (and 
the modification of it proposed by the 
Burgers 7 ) is based upon somewhat the 
same ideas of dislocation processes as seem 



macroscopically observable shear can re 
sult. If this requirement is actually opera 
tive, it throws some doubt on the validity 
of Taylor s theory, especially with regard 



POTENTIAL 
EMERGY 



CREEP RATE=C6 




SHEAR - 

FIG. 12. POTENTIAL-ENERGY CURVES ASSUMED IN BECKER S THEORY. 



to be required here, but the reason for 
strain-hardening in the light of Eq. r8 
requires further experimental work on the 
effects of the prestraining of single crystals 
on the values of AH*, AS*, and qAl. Tay- 
lor s theory amounts to an explanation of 
strain-hardening as the result of an increase 
in AH* alone, since, owing to the "attrac 
tions" of certain dislocations for one 
mother, it requires more energy to cause 
iislocations to progress when more dis- 
ocations appear in the metal 

It has been shown that there is an indi- 
:ation of a certain correlation between the 
jeparation of the "opaque surfaces" of 
Taylor and the lateral lengths of the dis- 
ocations. Unfortunately, it must be 
pointed out that Taylor s theory takes 
10 account of the requirement emphasized 
)n page 66 that in order for macro- 
;copically observable shear to occur, a 
lislocation must pass all of the way through 
.he specimen, so that if a dislocation is 
.topped at an opaque surface, no significant 



to the function of the opaque surfaces, so 
that thj correlation of these with the 
present theory is not as direct as it might 
seem. 

The theory of Becker 1 is very closely 
related to the present one, and could, 
indeed, very well be adjusted so as to take 
on the same form. Its essential characteris 
tic is that a definite shape is assigned to the 
potential-energy curve along which the 
unit flow process occurs. This shape is that 
given in Fig. 12, where the portions ABC 
and DBF have the form of parabolas, and 
the portion CD is a sharp cutoff of the 
parabolas, which corresponds to the shear 
at which the stress has its critical value. 
The advantage of this otherwise rather 
artificial assumption is that the difficulty 
of a possibility of a dependence of I in 
Eq. 1 8 on the applied force is avoided. It 
has been shown, however, that on theoreti 
cal grounds the effect would be expected 
to be small, while the near-linearity of log 



8o 



PLOW OF SOLID METALS AND CHEMICAL-RATE THEORY 



creep rate vs. stress plots shows that in 
general the effect is small. 

The interpretations of static tensile tests 
in the light of Becker s dynamical theory 
made by Orowan 5 are equally applicable 
with the present theory, and the value of 
regarding the problem of the plasticity of 
metals as essentially one of rates, which 
was emphasized by these writers, must 
not be overlooked. 

CONCLUSION 

If the question be asked, "What struc 
tural and chemical features of a material 
must be sought after in order that the 
material will creep at the slowest possible 
rate at a given temperature and under the 
action of a given stress?" the answer given 
by Eq. 18 must be that AS* should have as 
large a negative value as possible, AH* 
should be as large as possible, and qAl 
should be kept as small as possible. In 
the light of the results of our application 
of this equation to actual creep data, we 
must admit that the factors that tend to 
make AS* have a large negative value are 
as yet only incompletely understood. AH* 
can be made large by employing a material 
with a high elastic modulus, or by including 
in the structure of the material a network 
of chemical bonds with a strongly directive 
character instead of the less strongly 
directive metallic bonds,* or perhaps by 
the introduction of many dislocations in 
the manner of Taylor s theory. qAl can be 
made smaller by introducing more and 
more "opaque surfaces" or other barriers 



* This effect of directive bonds is prob 
ably responsible for the great resistance to 
creep shown by diamond, quartz, and many of 
the silicates on the one hand (all of these having 
strongly directive bonds), and for the very 
slight creep resistance of metallic sodium 
(which has a typically metallic bond). It may 
also be responsible for the greater hardness of 
steels over pure iron, the bonds in FesC being 
probably much more nearly homopolar than 
those in pure iron, and hence more strongly 
directed. 



preventing the lateral extension of dis 
locations (i.e., keeping A small). 

A fundamental attack on the problem of 
creep can be made by studying experi 
mentally and in more detail just what 
factors affect AS*, AH*, and qAl. If these 
factors were more thoroughly understood, 
it seems not unlikely that considerable 
advances could be expected in the develop 
ment of creep-resisting alloys. Experi 
mental work on creep in metals with a 
view to a more complete understanding 
of the applicability of Eq. 18 will begin 
shortly. 

SUMMARY 

1. Eyring s general statistical mechan 
ical theory of shear rates has been outlined 
and applied to available data on the creep 
of metals. 

2. From the values of the constants in 
Eyring s equation required to fit the data, 
it is shown to be very likely that metallic 
creep usually occurs by the shear of rather 
large blocks of material in the unit molec 
ular process, rather than by the movement 
of single atoms past one another as in 
diffusion. This is shown to fit in well with 
existing ideas on the flow of crystals by 
means of dislocations. 

3. The questions of why some metals are 
so much softer than others, the effects of 
impurities on hardness, the softness of 
single crystals, and the role of self -diffusion 
in creep are considered in the light of the 
theory, and it is shown how answers to 
these questions (and others similar) might 
be obtained by use of the theory. 

4. The relationship between the present 
theory and previous theories is discussed. 

ACKNOWLEDGMENTS 

The author wishes to express his grati 
tude to Dr. E. U. Condon for pointing 
out the applicability of Eqs. 15 and 18 to 
metallic creep and for making available 
his own unpublished work on the problem; 
and to Prof. F. Seitz, Dr. A. Nadai and 
Dr. T. A. Read for helpful discussions. 



DISCUSSION 



81 



REFERENCES 

1. A. Nadai: Plasticity. New York, 1931. 

2. E. N. Andrade: Phil. Mag. (1934) *7 (vii), 497, 

698. 

H. Eyring: Jnl. Chem. Phys. (1936) 4, 283. 
R. Becker: Phys. Ztsch. (1925) 26, 919; Ztsch. 

tech. Phys. (1926) 7, 547. 
E. Orowan: Ztsch. Phys. (1934) 89, 605, 614, 634; 

(I93S) 97, 573; (1936) 98, 382. 
G. I. Taylor: Proc. Roy. Soc. (1934) I45-A, 363, 

389, 411; Ztsch. Kristallographie (1934) 89, 383. 
W. G. Burgers and J. M. Burgers: First Report 

on Viscosity and Plasticity. Amsterdam, 1939. 
J. J. Kanter: Trans. A.I.M.E. (1938) 131, 385. 
E. U. Condon: Trans. A.I.M.E. (1938) 131, 410. 
R. H. Fowler: Statistical Mechanics, 30 ff. New 

York, 1936. 
R. C. Tolman: Statistical Mechanics, 71 ff. 

Oxford, 1938. 

E. U. Condon: Phys. Rev. (1038) 54, 957- 

W. Kauzmann and H. Eyring: Jnl. Amer. Chem. 

Soc. (1940) 62, 3113. 
H. Eyring: Jnl. Chem. Phys. (1935) 3, 107; Chem. 

Rev- (1935) 17, 65. 
M. G. Evans and M. Polanyi: Trans. Faraday 

Soc. (I93S) 31, 875. 
R. H. Ewell and H. Eyring: Jnl. Chem. Phys. 

Ci937) 5, 726. 
J. Hirschfield, D. Stevenson and H. Eyring: Jnl. 

Chem. Phys. (1937) 5, 907. 
R. H. Exvell: Jnl. Appl. Phys. (1938) 9, 252. 
D. Frisch, H. Eyring and J. Zincaid: Jnl. Appl. 

Phys. (1940) n, 75- 
C. F. Elam: Distortion of Metal Crystals, 34, 39. 

Oxford, 1935. 

A. Nadai: Jnl. Appl. Phys. (1937) 8, 418. 

B. Chalmers: Proc. Roy. Soc. (1936) I56-A, 427. 
J. McKeown: Jnl. Inst. Metals (1937) 60, 20. 

L. C. Tyte: Proc. Phys. Soc. London (1939) Si, 
203. 

A. A, Smith, Jr.: This volume, p. 165. 

L. C. Tyte: Proc. Phys. Soc. London (1938) So, 
159. 

W. M. Peirce and E. A. Anderson: Trans. 
A.I.M.E. (1929) 83, 560. 

H. Tapsell, A. E. Johnson and W. Clemshaw, as 
given by J. J. Kanter: Trans. A I.M.E. (1938) 
131, 385, Fig. 6. 

Creep Data, 1938. Compilation of Available High- 
temperature Creep Characteristics of Metals 
and Alloys, Amer. Soc. Mech. Engrs. and 
Amer. Soc. Test. Mat. (1938). 

F. L. Everett: Trans. Amer. Soc. Mech. Engrs. 
(i93i) S3, APM 53-10, 117, Figs. 18 and 19. 

H. Tapsell and A. E. Johnson, as given by J. J. 
Kanter: Trans. A.I.M.E. (1938) 131, 385, 
Fig. 3. Evaluation of constants by E. U. 
Condon. 

J. Gr6h and G. V. Hevesey: Ann. Phys. (1920) 



9. 
10. 



12. 
13- 

14. 
IS- 

16. 

17. 

18. 
19- 



21. 
22. 
23- 

24- 

25. 
26. 

27. 
28. 

29. 



30. 
31. 

32. 
33- 

34- 

35- 
36. 



[4] 63, 85. 

Landolt-Bornstein: Physikalische-Chemische Ta- 

bellen, Erg. Ila, 127; Erg. Ilia, 176. 
W. Seith: Diffusion in Metallen, 42-43. Berlin, 

D. Hanson: Trans. A.I.M.E. (1939) 133, 15- 
Hanfstengel and Hanemann: Ztsch. Metallkunde 

(1938) 30, 42, Fig. 3. 
H. F. Moore and B. B. Betty: Trans. Amer. Soc. 

Metals (1936) 24, 913. 



KEY TO NOTATION 

= shear rate. 

= distance in shear direction moved by 

flow units relative to one another in 

the unit process. See Fig. 2. 
= distance between layers of units of 

flow. See Fig. 2. 
net number jumps per second of unit 

of flow in shear direction. 
- energy. 



Af * = energy of activation per mol. 
AS* = entropy of activation per mol. 
AF* = free energy of activation per mol. 
k = Boltzmann s constant = 1.38 Xio" 16 

erg per molecule per degree C. 
h = Planck s constant = 6.626 X io~ 27 

erg-second. 
R = gas constant = 1.986 cal per mol per 

degree C. 

T = absolute temperature. 
A = area of unit of flow in the shear plane 

(area on which the shear force acts). 
AT" = Avogadro s number. 
5 = entropy. 

M Nm = molecular weight. 
I = distance through which the shear 

stress acts in carrying the unit of flow 

from the normal to the activated 

state. 

/ = shear stress acting on a unit of flow. 
a- = macroscopic ally observed shear stress. 
q = stress concentration factor = //tr. 
d = length of a dislocation. 
77 = coefficient of viscosity = cr/s. 
e energy of a molecule. 
m = mass of molecule. 
g = acceleration of gravity 

DISCUSSION 

(/. J. Kanter presiding) 

S. DUSHMAN,* Schenectady, N. Y. Inde 
pendently of Dr. Kauzmann, I had deduced 
from Eyring s theory an equation for creep that 
is in formal agreement with that derived in 
this paper in Eqs. 15 and 18. In view of the 
fact that the factor (\/L)(kT/h) in the latter 
has the dimensions of a frequency, the assump 
tion has been introduced that this frequency is 
determined by the Debye specific heat fre 
quency, which in iron, nickel and cobalt has 
the approximate value 8 X io 12 . Inserting this 
value into Eq. 18, it is thus possible to deter 
mine the free energy as a function of stress. 
Assuming, furthermore, that the change in 
entropy is zero, these considerations lead to a 
relation for creep of the form 



log 5 = log VQ 



<5o - /<*/ 
4-S78T 



[37] 



where the log is taken to base io, j/ is the 
Debye frequency, a is a constant which has 
the dimensions of a volume, / is the stress 
expressed in dynes per sq. cm., / is the recip 
rocal of the mechanical equivalent of heat; 
Qo is expressed in terms of calories per gram- 
atom, and T is the absolute temperature in 

* Research Laboratory, General Electric Co. 



82 



FLOW OF SOLID METALS AND CHEMICAL-RATE THEORY 



degrees Kelvin. Of course the creep rate must 
be expressed in terms of reciprocal seconds. 

In order to test the validity of this equation, 
an experimental technique has been used that 
may be described briefly as follows: The 
materials, such as 18-8, or other alloy, is 
drawn down to a wire of lo-mil diameter, and 
made up into V-shaped filaments 15 cm. long 
per leg of the V. This is inserted in a bottle 
through which purified nitrogen is passed 
slowly, and the weight is suspended from the 
bottom of the V. For the cross section of the 
wire used, a weight of i gram corresponds to a 
stress of 14 Ib. per sq. in. The filament is heat 
ed by passing current through it and is main 
tained at the desired temperature by comparing 
the brightness with that of a calibrated gas- 
filled tungsten lamp of the movie-projector 
type. Before the test of creep rate the filaments 
are annealed at a high temperature, without 
stress. 

While the results of the first experiments 
(on 1 8 Cr, 8 Ni alloys) seemed to be in agree 
ment with the assumption of zero entropy 
change, and this result was reported at the 
time of the presentation of this paper, sub 
sequent results have shown the necessity for 
introducing a finite entropy term, as has been 
done by Dr. Kauzmann. The magnitude of 
AS* varies from about 9 cal. per deg. per mol 
for 1 8-8 to about 56 cal. per deg. per mol for 
platinum. The corresponding values of a. 
and Qo for platinum are 75 and 44,000 cal. 
per gram-atom, respectively; while for 18-8 
these values are 31 and 97,000, respectively. 

The values of a. obtained in these experi 
ments are considerably greater than the 
atomic volumes, the ratio ranging from 100 to 
300. This may be due to a stress-concentration 
or the presence of atomic "domains" which 
move as units. The latter suggestion has also 
been made by Dr. Kauzmann. 

A more detailed discussion of the results 
obtained for a number of metals and alloys will 
be published in the near future. 

J. J. KANTER,* Chicago, 111. -It is gratifying 
to see the publication of a paper that seriously 
attempts to interpret creep phenomena from 
the viewpoint of physics and chemistry. The 
engineering application of creep data has 
encompassed little more analysis of creep and 

* Crane Co. 



flow phenomena than the calculation of 
empirical expressions useful in problems of 
applied mechanics. From the metallurgical 
point of view, however, more fundamental 
analyses of creep and flow will unquestionably 
prove of practical importance. Dr. Kauzmann s 
method of analysis affords a new approach 
toward this interpretion and merits thought 
ful consideration and study. It is to be hoped 
that experiments will be initiated specifically 
planned to test out the hypotheses of Kauz 
mann s analysis. 

There is one question with regard to the 
activation energies shown in Table i. Values 
are shown derived from creep tests for seven 
different ferritic steels ranging from 17,000 to 
102,000 cal. per gram mol. There is no very 
basic metallurgical difference among materials; 
they might have various grain sizes and other 
structural differences, but all these steels are 
composed of essentially the same type of 
ferrite crystals. It would not seem that a 
variation of stress concentration factor could be 
present among the steel to so widely vary the 
activation energy associated with slip. My 
question is in regard to the explanation of this 
wide divergence in the activation energies 
calculated according to Dr. Kauzmann s 
method upon similar materials. Does he regard 
the data used as so highly uncertain, or does 
this 600 per cent range of activation energy 
value represent an a tual metallurgical varia 
tion in the material? 

Another aspect of creep on which Dr. 
Kauzmann s interpretation would be of interest 
is the fact that such tremendously different 
creep rates are observed in a given material 
when its grain size is altered either through 
working or heat-treatment. The implications 
of Dr. Kauzmann s method of analysis should 
be studied in regard to such effects. 

T. A. READ,* East Pittsburgh, Pa In com 
paring creep data with Eq. 25, the author 
assumes that the dependence of the creep rate 
on stress at constant temperature is given 
entirely by the coefficient exp(qAlr/2KT). The 
success of this assumption is somewhat sur 
prising in the case of data on steady-state 
creep. One might expect the parameters AS* 
and A#* to depend on the stress, inasmuch as 

* Westinghouse Research Fellow, Westinghouse 
Research Laboratories. 



DISCUSSION 



they must be structure-sensitive properties. 
The structure of the metal presumably will 
be affected by prolonged loading, particularly 
at the higher stress levels. 

W. KAUZMANN (author s reply). It is inter 
esting that apparent confirmation of the 
general features of creep as given here has 
been found by Dr. Dushman in his experi 
ments, and the publication of his complete 
results is awaited with interest. It should be 
strongly emphasized, though, that owing to the 
frequent dependence of qAl, or a, on the 
temperature, reliable activation energies usu 
ally cannot be obtained from the temperature 
coefficient of the creep rate at constant stress. 
Indeed, this temperature sensitivity of the 
size of the unit of flow seems to be one of the 
outstanding characteristics of the whole phe 
nomenon of creep and demands a prominent 
position in any attempt at understanding 
creep. 

If the size of the unit of flow changes with 
temperature, the activation energy for their 
flow probably also changes with the tempera 
ture, and this bears strongly on Dr. Kanter s 
first point. If this temperature variation does 
occur, the method of obtaining the activation 
energies of Table i will not be valid. Therefore, 
as was pointed out in the paper, neither these 
values nor the entropies given in the table 
probably have physical significance as they 
stand. But if we are willing to say that the 
entropy of activation is zero (and this may 
not be far from the truth), the activation 
energy will be identical with the free energy of 
activation, AF* at any given temperature. 
A glance at Table i will show that ioooK. 



(- i3ooF., which is only a little higher than 
the temperature for which most of the data 
on the steels were obtained) the values of AF* 
for the steels are not very different and that, 
therefore, the true activation energies do not 
vary as markedly as Table i would indicate. 

It would not be altogether unreasonable, 
however, if there were a wide difference in the 
activation energies for creep between different 
steels: we must remember that there is every 
indication that structure as well as composition 
plays an important role in the plastic properties 
of metals, so that we must expect other factors 
beside the chemical composition to have their 
effect on the activation energy. 

The effects of grain size, heat-treatment, and 
cold-working on the creep rate from the stand 
point of the theory are now being investigated 
experimentally. These effects undoubtedly are 
complex, possibly involving simultaneous 
changes in AS*, A#* and qAl. 

Read suggests that by the time a metal has 
settled down to a steady creep rate at one 
stress, it might not have the same structure as 
when it is in the steady state corresponding 
to another stress, so that most of the data used 
here in applying the theory are not amenable 
to such an analysis. While it is important to 
keep this possibility in mind in evaluating the 
results of the analyses, the general validity of 
the exponential creep law for metals must lead 
us to conclude that things are really not so 
complicated. Besides, it is not too surprising 
that the stress, appearing here as it does in an 
exponent, should operate on the creep rate 
chiefly through this exponential function 
rather than indirectly through a second order 
effect on one of the other terms. 



Deformation and Recrystallization of Copper and 

Brass Hardness Microstructure and 

Texture Changes 

BY R. M. BRICK,* MEMBER, AND M. A. WILLIAMSON, f STUDENT ASSOCIATE A.I.M.E. 

(New York Meeting, February 1941) 



CERTAIN features of the response of 
copper and brass to deformation and 
recrystallization remain obscure. The tex 
tures obtained on rolled sheet are listed by 
Schmid and Boas 1 as: 



Metal 


Rolled 


Recrys- 
tallized 


Copper 


( I. (lIO)[lI2] 


(ioo)[iooj 


Brass 


(2. (ii2)Fiiij 

(l I0)[l I2J 


(i I3)[i i 2] 









No adequate explanation has been ad 
vanced for the presence of the secondary 
deformation texture of copper and its 
absence in brass or for the origin of the 
cubic recrystallization texture of copper. 2 
A recent publication 3 attempted to ration 
alize the above empirically described 
annealing texture of brass on the basis of 
retention of the (no) [112] rolling orienta 
tion plus twinning about the central 
octahedral poles of the usual secondary 
(no) [115] orientation. The difference in 
the textures of annealed copper as com 
pared to 70-30 brass is also shown by the 
difference in the position of ears for drawn 
copper, at o and 90, as compared with 
drawn brasses, at 45, for all compositions 
between QO-IO and 70-30. 4 Data on indi- 

A part of the work described in this paper has been 
presented in the form of a thesis by Merritt A. 
Williamson in partial fulfillment of the requirements 
for the degree of Master of Science at Yale University 
Manuscript received at the office of the Institute 
Nov. 25, 1940. Issued as T.P. 1299 in METALS 
TECHNOLOGY, February 1941. 

* Instructor in Metallurgy, Yale University, New 
Haven, Conn. 

t Metallurgist, Scovill Manufacturing Co., Water- 
bury, Conn. 

1 References are listed at the end of the paper. 



vidual crystals are as yet too few to permit 
of an analysis of these textures comparable 
to that of mild steel as made by Barrett and 
Levenson. 5 

EXPERIMENTAL METHODS 

The materials studied included copper 
specimens cut from the central, equiaxed 
crystal portion of a tough-pitch wirebar; 
brass made from this copper and high- 
purity (99.99 per cent) zinc including 
alloys of 7 per cent, 15 per cent, 24 per 
cent, 29 per cent and 36 per cent Zn and 
large single crystals of 70-30 brass. Sections 
% in. thick of the cast polycrystalline 
copper and brasses were reduced 40 per 
cent by rolling, recrystallized to a common 
grain size of 0.045 m m. and then rolled 
directly to foil 0.002 in. thick. Hardness 
readings employing both the Vickers and 
Rockwell machines were taken at intervals 
during the reduction. For comparison, 
similar specimens were cross-rolled i.e., 
turned through 90 for each pass and 
comparable hardness data were secured. 
Finally sections from the single crystals of 
brass were cut to obtain the following 
initial orientations; these sections were 
rolled, hardness data secured and a com 
plete micrographic record of structural 
changes obtained. 



Crystal 


A 


B 


C 


D 


Rolling plane. 


(100) 


(100) 


(no) 


(no) 


Rolling direction 


[100] 


Fnol 


flld 


Fnil 













84 



R. M. BRICK AND M. A. WILLIAMSON 



Crystal 


Fa 


G 


H 


I 


Rolling plane .... 


(112) 


(112) 


(in) 


(in) 


Rollins direction 


[ill] 


[no] 


[112] 


[no] 













a Textures were determined for these specimens 
at 80 and 99 per cent reductions. All pole figures 3 
were derived frorn^ X-ray photograms of 0.002-in. 
sheet taken at 10 intervals of rotation about longi 
tudinal and transverse axes. 



STRAIGHT AND CROSS-ROLLED COPPER 

Averages of the work-hardening curves 
of six different copper specimens are repro 
duced in Fig. i a. At all reductions between 
30 and 90 per cent, the straight-rolled 
copper showed a distinctly higher hardness 
than the cross-rolled copper, amounting to 
a maximum of 7 points on the Rockwell B 
scale. The difference was reproduced under 
varied rolling techniques employing 4-in. 
laboratory rolls but was not found when 
a large slab was reduced, in fewer passes, 
in a commercial mill employing large 
(i8-in.) rolls. 

The straight-rolled copper had an orien 
tation (Fig. 2 a) similar to that shown by 
Goler and Sachs 2 with one important 
exception, the as-rolled (99 per cent) sheet 
in addition to the usual (no) [112] texture 
and the secondary (112) [in] texture 
showed distinct evidence of material in the 
(100) [100] position. This is significant in 
view of the recrystallization texture (Fig. 
2 b} which, in agreement with other workers, 
was almost entirely in the cubic position; 
indeed the X-ray photograms appeared to 
show almost a single-crystal orientation. 
Thus copper conforms to many other cubic 
metals e.g., the Fe-Ni alloy studied by 
Sachs and Spretnak 6 inasmuch as a minor 
part of the deformation texture may 
become the chief or sole recrystallization 
orientation. This appears to be a more 
logical explanation of the cubic recrystal 
lization texture of copper than that offered 
by Cook and Richards 7 who suggest that 
the four central octahedral poles of the 
(no) [112] worked texture maintain their 
identity but adjust their positions. Such 



an adjustment would require rotations in 
mutually irreconcilable directions. 

The cross-rolled copper (Fig. 34), rather 
than showing three disparate orientations 
in the worked structure, appeared to be 
primarily in a (no) [223] texture, with 
other strong areas most readily explained 
on the basis of twinning about central 
(in) poles of the (no) [223] orientation. 
This cannot be construed as direct proof of 
mechanical twinning, which Mathewson 
and Van Horn 8 were not able to cause in 
copper. It is evidence that is suggestive of 
twinning, particularly in view of the lower 
hardness of the cross-rolled material. 
Twinning may be considered as an internal 
reorientation process tending to relieve 
crystallographic distortion, an indubitable 
source of hardness. Further support for this 
hypothesis is offered by the fact that the 
cross-rolled copper was found to have a 
slightly higher recrystallization tempera 
ture (Fig. i b) Sit 30 per cent, 50 per cent 
and 70 per cent reduction intervals (thus 
showing less internal distortion) and that 
for these specific specimens, 99 per cent 
reduction and J^ 1 hr. at 400 C. anneal, it 
showed a considerably coarser grain than 
the straight-rolled material. A complicated 
recrystallization texture was found and is 
explained in the pole figure (Fig. 36) as 
derived from the worked texture (110) 
[223], twins from this and from the weak 
(no) [112] deformation structure and also 
the usual (100) [100] copper recrystalliza 
tion texture. 

POLYCRYSTALLINE BRASSES, 7 TO 36 PER 
CENT ZINC 

The work-hardening capacity of solid 
solutions with varying solute content are 
part of a separate line of work and will not 
be reproduced here. However, in contrast 
to copper, no distinct and regular difference 
was found between straight-rolled and 
cross-rolled brasses of the same zinc con 
tent. The textures of all these normally 
rolled brasses was the same as that previ- 



86 



DEFORMATION AND RECRYSTALLIZATION OF COPPER AND BRASS 



ously reported for the 70-30 brass. 3 The 
deformation orientations were (no) [112] 
with a secondary spread (no) [113] [117]. 
The recrystallization textures obtained 



stresses of the cut, polished and etched 
faces. Furthermore, the Vickers impressions 
all showed long and short axes as well as 
convex or concave edges as a result of 



| 120 



100 



80 



66 



~* 40 



-6S 



=5 



o: 



Straight Rolled 
Ko/ltd 



SO 70 

PerCent Reduction 



90 



200 400 600 C 

Annealing Temperature 

a b 

FIG. i. WORK-HARDENING AND TYPICAL ANNEALING CURVES FOR STRAIGHT AND CROSS-ROLLED 

COPPER. 
b, 50 per cent reduction, ^-hour cumulative anneals. 



could again be empirically described as 
(113) [112] or more rationally as partly 
reproduction of the primary deformation 
texture plus twins from central octahedral 
poles of the secondary deformation orienta 
tion. The origin of the marked difference 
between textures and, relatedly, earing 
characteristics, of copper and brass has 
been located by Dahl 9 at between i and 5 
per cent Zn content, while considerably 
earlier work by Bauer, v. Goler and Sachs 10 
has shown that the transition in behavior 
occurs for alloys containing 1.8 per cent Zn. 

SINGLE CRYSTALS OF 70-30 BRASS 

Hardness values on the different crystal 
faces of these undeformed specimens were 
somewhat variable because of slightly 
different densities and the impossibility of 
assuring equal freedom from machining 



elastic and plastic anisotropy. Average 
readings gave relative hardness in the 
following decreasing order; (in), (no), 
(112) and (100), (221). 

The relative work-hardening rates are 
reproduced in Fig. 4. It appears that 
hardening takes place at a different rate 
and to a different degree as a function of 
the initial orientation, thus confirming 
previous studies performed at Hammond 
Laboratory. 11 Further work is required 
before these different rates can be directly 
related to crystal rotation as a unit or 
fragmentation with individual rotations 
into textural positions. 12 

Microscopic examination of the struc 
tural alterations during rolling included 
taking a large number of photomicrographs, 
most of which need not be reproduced 
here. 13 Certain generalizations may be 



R. M. BRICK AND M. A. WILLIAMSON 



stated on the basis of this evidence as 
follows: 

i. As in polycrystalline brass, no 
deformation markings appeared until re 
ductions exceeded 20 per cent. 



become transverse to the rolling direction. 
At this same reduction, the gross orienta 
tion has generally been found to start 
changing from essentially single positions 
to textural spreads, which in most cases 




a. Reduced 99 per cent. , b. Annealed y% hr. at 4ooC. 

FIG. 2. OCTAHEDRAL POLE FIGURES FOR STRAIGHT-ROLLED COPPER. 




a. Reduced 99 per cent. b. Annealed K hr. at 4ooC. 
FIG. 3. OCTAHEDRAL POLE FIGURES FOR CROSS-ROLLED COPPER. 

2. At low reductions, 20 to 40 per cent, could account for the markings as (in) 
the markings could be directly identified traces, but without definite assurance. 

as occupying octahedral planes of the One noteworthy exception to this vague 

matrix, identification at higher reductions was 

3. At higher reductions, about 70 per crystal H, rolled on a (in) plane in a 
cent, the deformation lines appeared wavy, [112] direction. At an 80 per cent reduction, 
broken up, branched but tending to it showed essentially a single (no) [100] 



DEFORMATION AND RECRYSTALLIZATION OF COPPER AND BRASS 



orientation (Fig. 5$), previously shown to 
be a fairly stable position up to reductions 
exceeding go per cent. Fig. 56 contains 
some straight-line markings whose direc- 



octahedral poles. It is significant in this 
connection to refer to Fig. 4^, which shows 
that this crystal did not harden, in fact 
softened, while being rolled from 50 to 



Hardness 



Numera 

a 



5 

^ 

<c 



/non 

S 



I 



3- 




ZO 




V / 

,o r^H-(in][iizj 




40 



40 



60 



60 20 

PerCent Reduction 

FIG. 4. WORK-HARDENING CURVES FOR SINGLE CRYSTALS or BRASS ROLLED FROM INDICATED 

INITIAL ORIENTATIONS. 



tions correspond to positions of the 
peripheral octahedral poles. A transverse 
photomicrograph confirmed this identifica 
tion of the markings as being traces of 



70 per cent total reduction. Fig. 5^ shows 
the shift to a nearly symmetrical normal 
deformation texture (no) [112] after 99 per 
cent reduction. The recrystallization tex 



R. M. BRICK AND M. A. WILLIAMSON 



8 9 



ture, not reproduced here, again appeared banded structure than Fig. 56. The 

to be derived from the worked structure recrystallization texture, Fig. 6b, showed 

directly or by twinning. six moderately strong peripheral (in) 

One other rolled single crystal in the areas, apparently directly derived from the 




a. Reduced 80 per cent. 



b. Comparable photomicrograph. X 100. 




A {110} [1/2] 



c. Reduced 99 per cent. 
FIG. 5. OCTAHEDRAL POLE FIGURES OF SINGLE CRYSTAL (H) OF BRASS ROLLED FROM AN INITIAL 

(ill) [lI2] ORIENTATION. 



initial orientation (112) [in], was studied 
on the basis of its deformation and re- 
crystallization textures. After a reduction of 
80 per cent, the material was still near its 
original orientation but, in addition, had 
developed in equal intensity a complete 
mirror image of this (Fig. 60) corresponding 
to a much heavier and also less well defined 



deformation texture. It is interesting that 
this conformation has been found in com 
mercial brass with the corresponding 
development of six ears on drawn cups.* 
The same crystal after a 99 per cent reduc 
tion (Fig. 6c) again approached the normal 
rolled texture (no) [112] and the texture of 

* Private communication; Dr. H. L. Burghoff. 



DEFORMATION AND RECRYSTALLIZATION OF COPPER AND BRASS 



the recrystallized specimens Fig. 6d ap 
proached the normal annealing orientation, 
empirically (113) [112]. It is particularly 
evident that areas marked A in Fig. 6d 

a. Reduced So per cent. 



SUMMARY 

Copper cold-rolled in one direction work- 
hardened more than comparable cross- 
5. Annealed }i hr. 4ooC. 




c. Reduced 99 per cent. d. Annealed l / 2 hr. 4ooC. 

FIG. 6. OCTAHEDRAL POLE FIGURES OF SINGLE CRYSTAL (?) OF BRASS ROLLED FROM AN INITIAL 

(lI2) [in] ORIENTATION. 



cannot be readily attributed to reproduc 
tion of the rolling texture but do conform to 
the empirical (113) [112] orientation or 
to twins of the rolled structure. It appears 
to be significant in view of the fact that 
this specimen had a work-hardening curve 
(Fig. 40) more like that of polycrystalline 
brass than any of the other single crystals. 



rolled metal. The straight-rolled material 
had a texture similar to that described in 
the literature, both (no) [112] and (112) 
[in] structures being present and in addi 
tion some material in the hitherto unre- 
ported (100) [100] orientation. The latter 
is significant since it constitutes the known 
recrystallization texture, which in this case 



DISCUSSION 



was so sharp that the material appeared to 
be nearly a single crystal. The recrystalliza- 
tion process thus appears to consist 
primarily of the growth of stable (100) [100] 
nuclei at the expense of the dominant but 
less stable other orientations. 

The lower hardening of the cross-rolled 
copper may plausibly be attributed to 
mechanical twinning acting to relieve 
internal distortion since the texture of this 
sheet was best explained by a (no) [223] 
orientation plus twins from the central 
octahedral poles of the texture. This 
hypothesis is also consistent with the 
observed higher recrystallization tempera 
ture and coarser grain of the cross-rolled 
material. The recrystallization texture was 
in this case very complicated but appar 
ently consisted of the original orientations 
in somewhat shifted relative intensities. 

Polycrystalline alpha brasses of from 
7 to 36 per cent Zn were quite different 
from copper and corresponded, texturally, 
to the 70-30 brass previously reported. 

Several single crystals of brass were 
rolled and the accompanying changes in 
their hardness, microstructures and in two 
cases, textures, were determined. The data 
were in conformity with previous work in 
showing that deformation markings at low 
reductions were along octahedral planes. 
In one specimen rolled to a 70 per cent 
reduction, the pole figure rather definitely 
indicated its deformation lines occurred 
along (in) planes and thus could be 
mechanical twins. This same specimen 
significantly failed to work-harden on being 
rolled from 50 to 70 per cent reduction. 
The end deformation textures again 
approached the normal (no) [112] struc 
ture although other relatively stable orien 
tations appeared to persist beyond an 
80 per cent reduction. The symmetric or 
nonsymmetric recrystallization textures 
could best be rationalized on the basis of 
retention of original orientations or twins 
therefrom. 



REFERENCES 

1. E. Schmid and W. Boas: Kristallplastizitat. 

Berlin, 1935. Julius Springer. 

2. v. Goler and G. Sachs: Ztsch. Phys. (1927) 41 

873, 889; (1929) 56, 477, 485. 

3- R., M. Brick: Trans. A.I.M.E. (1940) 137, 193 

4- A. Phillips and C. H. Samans: Trans. A.I.M E 

(1933) 104, 171. 

5- C. S. Barrett and L. H. Levenson: A.I.M.E. 

Tech. Pub. 1233 (Met. Tech., Sept. 1940). 
6. G. Sachs and J. Spretnak: Trans. A.I.M.E (1940) 

140, 359- 
7- M. Cook and T. L. Richards: Jnl. Inst. Metals 

(1940) 66, i. 
C. H. Mathewson and K. R. Van Horn: Trans. 



A.I.M.E. (1930) 59- 
). Dahl and F. Pawlek: 
28, 266. 



9. 0. Dahl and ] 



Ztsch. Metallkunde (1936) 



10. 0. Bauer, F. Goler and G. Sachs: Ztsch. Met all- 

kunde (1928) 20, 202. 

11. M. R. Pickus and C. H. Mathewson: Trans 

A.I.M.E. (1939) 133, 161. 
C. H. Saraans: Thesis, Sterling Library, Yale 

University, 1934. 
H. L. Burghoff: Thesis, Sterling Library, Yale 

University, 1930. 

12. H. C. Vacher: Nat. Bur. Stds. Jnl. Research 

(1939) 22, 651. 

13- M. A. Williamson: Thesis, Sterling Library, 
Yale University, 1940 

DISCUSSION 

(/. T. Norton presiding) 

C. S. BARRETT,* Pittsburgh, Pa. An exten 
sive study of the textures in rolled copper and 
brass has been made by two Russians, 14 and 
should be considered in connection with these 
results. Pole figures for (in), (200) and (220) 
planes were accounted for by them as the sum 
of the "ideal orientations" shown in Table i. 

TABLE i. Ideal Orientations of Iweronowa 
and Schdanow 







R.D. 




Intensity 


Orienta 
tion No. 


R.P. 


Cop 
per 


Brass, 


Brass, 
So to 20 








95 to 5 


and 








Per Cent 


62 to 28 










Per Cent 


i 


(no) 


lI2 




S 


S 


S 


2 


(112) 


[III] 


S 






3 


(124) 


335 




M 


W 


W 


4 


(no) 


100 




W 


M 


S 


S 


(100) 


100 




W 







In straight-rolled copper they found all five 
orientations with the relative intensities 
(strong, medium, and weak) listed in the table; 
in the brasses ranging from 5 to 28 per cent Zn, 

* Metals Research Laboratory, Carnegie Institute 
of Technology. 

14 W. Iweronowa and G. Schdanow: Walztextur des 
Alpha- Messings. Tech. Phys. U.S.S.R. (1934) i. 64. 



DEFORMATION AND RE CRYSTALLIZATION OF COPPER AND BRASS 



the intensities were altered as shown, and some 
components disappeared (Nos. 2 and 5). 
Schmid and Boas found the first two ideal 
textures in copper, the first one only in brass; 
Sachs and Spretnak found Nos. i, 2 and 5 
in Fe-36 per cent Ni. Fig. 20 herein, for straight 
rolled copper, is interpreted in terms of Nos. 
i, 2 and 5, but 3 and 4 also fall within the 
shaded areas of the pole figure, just as in the 
very similar texture found by the Russians. 

It is always recognized that the details of 
the pole figures are not fully described by one 
or two ideal textures, and even with this list 
of five there are still significant details that 
are not indicated. There is the interesting fact, 
for instance, that in rolled brass there is 
considerable material with (in) in the rolling 
plane, while there is none in copper. Figs. 5 a, 
5c, 6a and 6c are in accord with this behavior 
of brass, for all show material at the center of 
the (in) pole figure. That the list is not com 
plete is suggested by reference to Dr. Brick s 
paper (ref . 3) on rolled brass, and also by some 
tests made at the Metals Research Laboratory 
on copper crystals reduced 96 per cent by 
rolling, which have taken orientations other 
than any on this list, yet lying within the 
dense areas of the polycrystalline pole figure. 
Probably, also, there are orientations in the 
list that are not stable at high reductions. 
Crystal H arrived at orientation No. 4 after 
80 per cent reduction, but gave it up in favor 
of No. i with further rolling. We have rolled 
some crystals of copper and have had multiple 
orientations form from a crystal in orientation 
No. 2. 

I should like to add a word about the role of 
twinning. There are 20 different twins of the 
five textures listed above; allowing for those 
that are similarly related to the rolling direction 
and rolling plane there are still, I believe, 13. 
These, together with the five orientations 
listed, give a total of 18. Very few orientations 



can be found that do not lie near one of these 
18; consequently, it is difficult to test a twinning 
theory by an investigation of the polycrystal 
line pole figure for twinning relationships 
within it. There are other difficulties also: a 
plane of symmetry of the rolling process may 
also be a twinning plane for one of the ideal 
textures. This is true of orientation No. i and 
weakens some of the conclusions about twin 
ning in the paper by Cook and Richards (ref. 7). 
We are rolling some copper crystals in an 
attempt to get more definite tests of the 
twinning theory and a better general under 
standing of the origin of the polycrystalline 
texture; thus far multiple orientations have 
developed in several crystals, but orientations 
that are twins of each other have not been 
found. 

R. M. BRICK (author s reply). We are 
grateful to Dr. Barrett for calling our attention 
to the work by the Russian authors. The com 
parisons and interpretative statements made 
by Dr. Barrett are in accord with our ideas and 
require no further comment. 

In regard to the twinning mechanism, it must 
be conceded, of course, that the pole figure of a 
severely deformed metal cannot prove that the 
twinning mechanism is operative or even 
demonstrate with certainty that twinned 
orientations are present. On the other hand, 
the pole figure of the cross-rolled copper fits in 
very nicely with the twinning hypothesis and, 
as Mr. Williamson pointed out in the presenta 
tion, we believe that a lesser work hardness 
would accompany twinning. Since it cannot be 
proved by the pole-figure method, the matter 
must remain at its present status. Perhaps 
other methods of attack will be more successful 
in demonstrating whether deformation mark 
ings in copper and brass are twinning, and if 
they are not, what is their function in the 
rolling and crystallization processes. 



Corrosion of Copper and Alpha Brass Film-structure Studies 



BY J. H. HOLLOMON/ 



SERVICE failures in 
tubes are often due to corrosion. One of the 
commonest types of corrosion reveals a 
surface structure of redeposited copper. 1 
The study of the effect of alloy addition 
agents in preventing this kind of corrosion 
has received impetus from the extensive 
publication of Bengough and May. 2 The 
mechanism of such inhibition gauged from 
immersion corrosion, 2 cell corrosion, 3 and 
electrolytic potential 4 measurements is 
controversial. This controversy may be 
attributed in part to a lack of definite 
knowledge 5 of the actual nature of the 
initial films formed in corrosion experi 
ments. In our own corrosion work on 
copper and 70-30 brass this lack be 
came apparent and necessitated an exami 
nation of surface films by X-ray and 
electron diffraction methods. The present 
paper is but an introduction to the subject 
of inhibition and reports the structure of 
films formed in active corrodents. 

EXPERIMENTAL PROCEDURE 

The arsenical coppers and brasses used 
in this investigation were made by one of 
the writers* in the plant of the Revere 
Copper and Brass Co. at Rome, N. Y. 
Standard commercial melting practice was 
followed. O.F.H.C. copper and electrolytic 
zinc were used and all arsenic additions 
were made by additions of a master copper 

Manuscript received at the office of the Institute 
Dec. 2, 1940. Issued as T.P. 1311 in METALS TECH 
NOLOGY, April 1941. 

* Research Assistant, Department of Metallurgy, 
Massachusetts Institute of Technology, Cambridge, 

t Associate Professor of Physical Metallurgy, 
Massachusetts Institute of Technology, Cambridge, 
Mass. 

i References are at the end of the paper. 



STUDENT ASSOCIATE, AND JOHN WuLFF,f MEMBER A.I.M.E. 

(New York Meeting, February 1941) 

brass condenser alloy. The final analyses are given in 
Table i. Likewise, the extrusion, cold- 
rolling, and intermediate anneals em 
ployed followed standard usage. Final heat- 
treatments were given to all materials to 
produce a similar grain size (0.035 to 
0.045 mm. for brass and 0.015 to 0.025 f r 



TABLE i. Analyses of Arsenical Brass and 
Copper Employed 



No. 


As, 
Per Cent 


Cu (O.F.H.C.), 

Per Cent 


Zn (by Dif 
ference) , 
Per Cent 


I 


nil 


69.46 


30.54 


2 


0.014 


69-40 


30.59- 


3 


0.024 


69.66 


30.32- 


4 


0.031 


69-42 


30.55 


5 


0.045 


69-70 


30.25 


6 


0.087 


69-51 


30.40 


7 


0.256 


69-74 


30.00 


8 


0.425 


69-51 


30.07 


9 


0. 10 


70.00 


30. 


Ci 


nil 


remainder O.F.H.C. 




C2 


0.013 






C3 


0.023 






C 4 


0.040 






Cs 


0.084 






C6 


o. 206 






C 7 


0.413 






C8 


0.603 







copper) for the alloys containing various 
percentages of arsenic. Samples % by }>i 
by Y in. were cut from sheet material 
resurfaced and finished with oooo metal- 
lographic papers. The surfaces were all 
degreased in alcohol and benzine before 
immersion. After immersion in the cor 
rodent, they were washed gently in tap 
water, dipped first in alcohol and then in 
pure benzine and allowed to dry in the 
diffraction apparatus. 

The reflection electron diffraction appa 
ratus used has been described by Burwell 
and Wulff. 6 Fig. i is typical of the photo- 
grams obtained. The X-ray apparatus was 



93 



94 CORROSION OF COPPER AND ALPHA BRASS FILM-STRUCTURE STUDIES 



of standard powder type employing Phrag- 
men cameras. After insertion of the sample 
in the electron diffraction chamber, about 
30 min. was required to evacuate the 




FIG. i. ELECTRON DIFFRACTION PATTERN OF 

O.F.H.C. COPPER CONTAINING 0.084 ARSENIC 
AFTER 40 MINUTES IN 2N HNOg. 

This pattern consists of rings of As20s and 
copper. 

chamber, and an exposure of i min. to a 
3o-kv. electron beam to procure a diffrac 
tion photogram. The utility of the electron 



diffraction method for film studies has been 
pointed out sufficiently elsewhere; 7 it per 
mits, in most cases, the study of films less 
than 100 A. thick, whereas nearly all 
X-ray diffraction beams penetrate too 
deeply into the substrate to be of value in 
studying very thin films. The results are 
given in Table 2. 

Nonarsenical coarse-grained 70-30 brass, 
when finished with oooo paper and thor 
oughly degreased, gives a pattern repre 
sentative of cold-worked alpha brass. The 
cold-working effects are due to surface 
finishing, since the bulk material is coarse 
grained. After being etched to a depth of 
about 5000 A., a spotted pattern character 
istic of the coarse-grained brass results. 
Specimens of nonarsenical brass when 
immersed in a loo-c.c. solution of 5N HN0 3 
for i hr. give a pattern of copper oxide 
(CuO) and coarse-grained brass. The CuO 
film is not visible to the eye for such short 
periods of immersion, although the grain 
structure of the basis material is micro 
scopically apparent. When the sample con- 



TABLE 2. Results of Electron Diffraction and Examination of Corrosion Films of Arsenical 

Brasses and Coppers 
Analyses of alloy content of samples given in Table i 



Pattern No. 


Sample 
No. 


Corrodent (Volume 
= TOO C.C.) in Open 
Beaker 


Time of 
Immersion 


Surface Film 


Remarks 


i 


i 


SNHNOs 


i hr. 


CuO -f- Br 


Film not visible 


2 

3 


4 
4 


SNHN0 3 
SNHNOa 


i min. 
30 min. 


CuO + As 2 3 -f Br 
As 2 0s 


Very thin 
Very thick 


4 


6 


sNHNOs 


2 min. 


CuO -f As 2 O 3 + Br 


Very thin 


S 


6 


$N HNOs 


48 hr. 


As 2 0s 


Very thick 


6 


9 


2NHNO 3 


10 min. 


Sb 


Thin 


7 


Ci 


SNHNOa 


i hr. 


Cu + CuO 


Not visible 


8 


C 4 


2NHNO 3 


20 min. 


As 2 0s + Cu 


Coarse As 2 0s 


9 


cs 


2N HNO 3 


20 min. 


As 2 3 + Cu 


Coarse As20s 


10 


C6 


2NHNOs 


20 min. 


As 2 0s -f- Cu 


Coarse AsaOs 


II 


i 


2N HC1 


6 hr. 


Brass 




12 


i 


2N HC1 


5 days 


Cu or brass 




13 


3 


2NHC1 


S days 


Cu or brass" 




14 


4 


2NHC1 


3 days 


Cu + As 2 0s 


Thin 


IS 


6 


SN HC1 


2hr. 


Cu + As 2 0s 


Thin 


16 


i 


i gram,As20s 


ihr. 




Metallic crystals 






in 500 c.c. HCK6N) 








17 


9 


2NHC1 


10 min. 


Sb 




18 


C6 


2NHC1 


I day 


Cu + As 2 3 


Thin 


19 


c? 


2NHC1 


i day 


Cu -f As 2 0s 


Thin 


20 


C4 


2NHC1 


6 days 


CuCl 


Precipitate 


21 


C4 


2NHC1 


6 days 


As 2 3 + CuCl 


No. 20 scraped 



In electron diffraction work it is extremely difficult to distinguish between fine-grained alpha brass and 
fine-grained copper, because of the slight effect of zinc on the lattice parameter. Nevertheless, coarse-grained 
alpha brass can be distinguished from fine-grained alpha brass or copper. When grain boundaries are micro 
scopically visible, it is reasonable to assume that all the fine-grained brass arising from the cold-work of surface 
finishing has been removed and any fine-grained diffraction pattern obtained may then be attributed to copper. 

* Auxiliary X-ray patterns show that the deposit contains metallic arsenic. 



J. H. HOLLOMON AND JOHN WULFF 



tains more than 0.03 per cent As, it is 
possible in i min. in 3 to 6N HNO 3 to 
procure a film that is just discernible, and 
whose diffraction pattern is that of copper 
oxide plus arsenious oxide (CuO + As 2 3 ), 
and coarse grained alpha brass. If the time 
of immersion in the same medium is 
increased, the film becomes visible to the 
eye. For a sample containing 0.045 P er 
cent As immersed in 5N" HNO 3 for 30 min., 
the top 100 A. of the film are essentially 
arsenious oxide (As 2 3 ). Furthermore, the 
As 2 O 3 formed in this length of time is much 
coarser grained than is the initial arsenious 
oxide film. Thick films of this kind are not 
very tenacious and when rubbed of! leave 
an adherent bare film of CuO + As 2 O 3 . 
Nonarsenical copper immersed in 5N HNOs 
for i hr. does not form a film visible to the 
eye, yet the electron diffraction pictures 
reveal a copper oxide coating. When 
arsenic in amounts greater than o 04 per 
cent is present in copper, the film consists 
essentially of As 2 3 . For exposures less 
than i hr. in 2N HNOs, and for much 
shorter times in more concentrated nitric 
acid, a coarse-grained arsenious oxide film 
results. For long periods of immersion in 
weak acid the copper oxide film cannot be 
detected. This does not mean that CuO 
is absent, but only that it is present in 
amounts less than 10 per cent in the mate 
rial that diffracts the electron beam. 

Of particular interest are samples of 
antimonial 70-30 brass, which, after 10 
min. in 2N HNOs, exhibit thin films whose 
structure analyses indicate the presence of 
metallic antimony as the sole component 
of the corrosion film. 

The foregoing results represent experi 
ments made in extremely oxidizing environ 
ments. Another set was performed in 
open beakers with various normalities of 
hydrochloric acid. From nonarsenical alpha 
brass after 5-day immersion in 2N HC1 a 
pattern of coarse-grained brass (or copper) 
was obtained. The films were invisible, 
yet the grain structure of the material was 
apparent under the microscope, indicating 



that corrosion had taken place. For longer 
immersions or in more concentrated acid 
when color changes were noted, fine-grained 
copper films typical of "dezincified" mate 
rial were found. 

When the thick porous arsenical film 
was scraped from an alpha brass immersed 
in a saturated solution of As A in 6N HC1, 
the underlying surface of brass was found 
to be covered with copper. If the coating 
is scraped a short time after immersion, 
this copper film is not discernible. An 
analogous experiment was performed by the 
immersion of a straight alpha brass in 
6N HC1, to which one gram of Sb 2 3 was 
added. Metallic antimony immediately 
deposited on the brass. When thin coatings 
were rubbed off, the underlying metal was 
of the original color. When the reaction was 
allowed to proceed for sufficient time (6 to 
12 hr.) to produce a thick porous coating, 
the underlying brass was found to be 
covered with bright copper. Evidently 
corrosion of the nonarsenical and non- 
antimonial brasses in this experiment 
proceeds through the pores of the arsenical 
and the antimonial film, thereby causing 
" dezincification " of the brass. 

Arsenical coppers containing more than 
0.04 per cent As, when immersed in 2~N HC1 
for times of the order of 24 hr., are covered 
with extremely thin films primarily of 
As 2 3 . If, instead of immersion in 100 c.c. 
of solution, 25 c.c. is used just to cover the 
samples, the solution becomes cloudy after 
6 days. In order to distinguish between 
precipitation and electrolytic deposition 
of corrosion products, the structure of the 
extremely thick film formed was examined. 
Between the top CuCl film and the copper 
a porous arsenious oxide (AsaOJ film was 
found. 

DISCUSSION OF RESULTS AND CONCLUSIONS 

In the corrosion of copper it is evident 
that a crystalline film of copper oxide 
forms on the surface in certain corrodents. 
When the copper contains arsenic, redeposi- 
tion of metallic arsenic takes place which, 



9 6 



CORROSION OF COPPER AND ALPHA BRASS FILM-STRUCTURE STUDIES 



in oxidizing media, is rapidly converted to 
arsenious oxide (AsaOs). When immersed 
in hydrochloric acid solutions open to the 
air, the corrosion process is extremely slow 
and the arsenical film found upon the 
surface is identical with that found in 
oxidizing media. 

Brasses containing less than 0.024 to 
0.03 per cent As when immersed in nitric 
acid are filmed primarily with CuO, which 
may slow down corrosion but does not 
prevent it. For higher arsenic contents the 
brass corrodes as an entity and copper 
redeposits and is oxidized. Since the 
corrodent rapidly becomes richer in arsenic, 
this element begins to redeposit and is 
oxidized to As 2 C>3. For the antimonial brass 
studied, the mechanism of deposition is 
essentially similar, yet redeposited anti 
mony does not oxidize as readily. In every 
case studied the initial film formed is more 
tenacious than later films, but in none does 
the film appear impervious to the corrodent. 

For brasses in hydrochloric acid of dif 
ferent normality, the first film formed is 
probably copper. When such brasses con 
tain more than about 0.02 to 0.03 per cent 
As, the upper part of the film is essentially 
arsenious oxide (AsgOs). For equivalent 
times in hydrochloric acid the film is 
extremely thin as compared with that 
formed in HNOg. Although little oxygen is 
available, arsenic apparently is deposited 
as such but is oxidized to arsenious oxide. 
The chemistry of this is in harmony with 
the reactions of metallic arsenic stated by 
Prescott and Johnson 8 for solutions up to 
6N HCL 

Reflection X-ray experiments show that 
metallic arsenic readily deposits on alpha 
brass from saturated solutions of As20s in 
2N to 6NHC1. Electron diffraction pic 
tures of this deposit show that it is also 
covered with a film of As20s, therefore it 
seems reasonable to conclude that during 
the corrosion of arsenical brass the initial 
deposit is also metallic but is oxidized when 
sufficient oxygen is available. We do not 



believe that the oxide film serves subse 
quently as a cathode; the corrosion process 
continues albeit more slowly since the 
film has some porosity, otherwise the film 
would not increase in thickness with time. 
When the metallic arsenic film deposited 
on nonarsenical brass immersed in the 
HCl-As 2 O 3 solution is gently rubbed off 
with cotton, the basis film is alpha brass. 
As deposition is continued, the basis film 
becomes more copperlike in color and 
structure, indicating that dezincification 
has taken place not only through the pores 
of the arsenious oxide but also through the 
pores of the metallic arsenic film. 

A similar mechanism exists for anti 
monial brass, as indicated by the immer 
sion of ordinary brass in a solution of 
Sb 2 O 3 in 6NHC1. The absence of the 
observable oxide film in the antimonial 
brass immersed in hydrochloric acid is 
probably not contradictory. The porous 
antimony film formed does not necessarily 
function electrolytically in the corrosion 
process. 

The information obtained on the struc 
ture of corrosion films of arsenical brass, 
arsenical copper and antimonial brass 
indicates: (i) that the conclusions of 
Masing 4 based on overpotential studies of 
antimonial brasses need revision, and (2) 
that Fink s 3 viewpoint on the dezincifica 
tion of arsenical brass requires extension. 

ACKNOWLEDGMENT 

The present work has been made possible 
by a research grant from the Revere Copper 
and Brass Company. 

REFERENCES 

1. R. B. Abrams: Trans. Amer. Electrochem. Soc. 

(1928) 12, 39. 

2. G. E. Bengough and R. May: Jnl. Inst. Metals 

(1924) 32, 81. 

3. F. W. Fink: Trans. Electrochem. Soc. (i939) 75, 

44I-. 

4. G. Masing: Wiss. Ver. Siemens (1931) 10, 2. 

5. F. R. Morral: Trans. Electrochem. Soc. (1939) 75, 

448. 

6. J. Burwell and J. Wulff: Trans. A.I.M.E. (1939) 

135, 486. 

7. G. I. Finch and H. Wilman: E. E. Naturw. (1937) 

l6 353- 

8. Prescott and Johnson: Outline of Chemical 

Analyses, 194. New York, 1933. 



DISCUSSION 



97 



(L. L. Wyman presidi-, 

H. R. HANLEY,* Rolla, Mo. My interest in 
this paper is not particularly that of corrosion, 
but I am interested in the film of arsenic. It is 
totally unrelated and perhaps my remarks are 
not apropos here. But I do notice that when 
steel is pickled in hydrochloric solution, there 
is no hydrogen embrittlement, but when the 
steel is pickled in sulphuric acid there is such 
embrittlement. I have been told that perhaps 
an arsenic film prevents the absorption of 
hydrogen atoms. 

N. W. MITCHELL, t "Waterbury, Conn. I 
wonder whether the authors have obtained any 
data as yet to indicate just what effect the 
concentration ratio of copper to arsenic in the 
solution has on the rate, and also on the type of 
product deposited. 

F. R. MCXRRAL, J Kokomo, Ind. The authors 
mentioned that under certain conditions the 
electron diffraction pattern reveals copper 
oxide coating below the arsenious oxide film. 
For long periods of immersion in weak acid the 
copper oxide film cannot be detected. 

My experience with corrosion products on 
zinc coatings indicated that although by 
electron diffraction we obtained only a pattern 
of basic zinc carbonate, by X-rays we obtained 
the patterns of the layers underneath; namely, 
zinc oxide and zinc. Have the authors made 
both types of analysis of the sample? If they 
were able to strip the film, have they tried to 
obtain electron diffraction patterns on the metal 
side of the film? 

C. E. SWARTZ, Cleveland, Ohio. There is 
one point in which some of us are interested 
the use of metals in conditions where solutions 
do not come in contact with them. I am think 
ing now of nonaqueous types of corrosion, such 
as the corrosion obtained at elevated tempera 
tures in various types of atmospheres. Another 
type of corrosion that is very difficult to deal 
with is that where oils of various kinds come 
in contact with metal oils having dissolved in 
them or suspended in them certain corroding 
agents. The latter kind of corrosion is found in 
most of the internal-combustion engines, 
where the oil corrodes the bearings. 



* Missouri Scliool of Mines and Metallurgy. 

! Chase Brass and Copper Co. 
Continental Steel Co. 
Cleveland Graphite Bronze Co. 



J. H. HOLLOMON AND J. WuLFF (authors 
reply). Mr. Hanley s remarks are of interest 
to us for we have for some time been interested 
not only in the effect of arsenic, phosphorus, 
and antimony on pickling but also on other 
corrosion phenomena. Our corrosion experi 
ments indicate that, even if iron or steel 
contained arsenic to a more limited extent 
than that found in ordinary arsenical brass, 
arsenic would redeposit upon the surface of 
the metal during corrosion. Such redeposition 
would occur whether the alloy corroded by 
hydrogen evolution or by reduction of oxygen 
to hydroxyl ion. The hydrogen overvoltage 
of arsenic is considerably less than that of 
iron and it seems that the presence of an 
arsenical film would facilitate hydrogen 
removal at the surface. To our knowledge, 
there are no published data upon the rate of 
corrosion by hydrogen evolution as a function 
of the arsenic content in either iron or steel. 
It is possible that the corrosion rate of iron 
may be decreased by the addition of arsenic 
even though it allows the evolution of hydrogen 
at the surface. We have been investigating 
this point and hope to publish data upon it 
in the near future. 

In answer to Mr. Mitchell, we can only 
say that in a future paper we will include a 
discussion of the mechanism of the corrosion 
of arsenical brass. For the present, it can be 
said that a certain critical arsenic content 
is required in order to permit its redeposition. 
This critical composition depends primarily 
on the corrosive medium. After such deposi 
tion occurs, the nature of the film determines 
the concentration of cuprous ions necessary 
to allow the deposition of metallic copper. 

In answer to Mr. Morral: Supplementary 
X-ray diffraction experiments indicate that 
the underlying film consists of metallic arsenic. 
Further experiments are necessary, however, 
to ascertain definitely the structure of all the 
components of the underlying film. 

Mr. Swartz comment suggests an interesting 
possibility for some experimental work. The 
preliminary experiments carried out indicate 
that the arsenical alloys, when heated in air 
under 500 C., are filmed primarily with 
arsenious oxide. It is necessary to determine 
at what temperature this compound ceases 
to be the essential constituent and the oxides 
of copper become paramount, for arsenious 
oxide might have a tendency to sublime at 
elevated temperatures. 



Some Practical Observations on Inverse Segregation 



BY DANIEL R. HULL,* MEMBER A.I.M.E. 

(New York Meeting, February 1941) 



IN 1926 Genders 1 reviewed the existing 
theories of this subject and stated his 
views in support of the gas-pressure theory. 
Again, in 1937, the subject was thoroughly 
reviewed by N. B. Vaughan, 3 and by 
Phillips and Brick, 2 so that no resume is 
called for here. The purpose of this paper 
is to describe certain observations made 
in the course of mill work, and to point 
out that they mostly tend to confirm the 
gas theory of Genders. 

VAPOR SEGREGATION 

There is, however, one type of inverse 
segregation that can be accounted for 
by considering the volatility of one of 
the constituents of an alloy. The most 
familiar case is ordinary brass, which is 
subject to a zinc enrichment of a few tenths 
of a per cent at the surface of a chill 
casting. This has been noted in 70-30 brass 
by Genders, and a typical example of it 
is given in Table i, which represents a 

TABLE i . Example of Inverse Segregation 



Sample 


Copper, 
Per Cent 


Lead, 
Per Cent 


Surface 3^2 i n - 


6 1 .43 


3 18 


2nd layer.. . . . 
3rd. layer 


61.50 
6 1 42 


3-19 
3 32 


4th layer 


61 65 


32 t 


5th layer 


61.83 


3 18 


6th layer 


61 .89 


3 II 


Center 


61 86 


31 A 









cylindrical billet 7 in. in diameter by 
55 in. long. After lathe-turning J 2 in. 

Manuscript received at the office of the Institute 



Dec. 2, 1940. Issued as T.P. 1287 in METALS TECH 

Y, F 

etall 
bury, Conn. 



. 

, ebruary 1941. 
Metallurgist, The American Brass Co., Water- 



, 
1 References are at the end of the paper. 



from the surface, the remaining distance 
to the center was sampled at six equally 
spaced intervals. Tlie lead, which is essen 
tially the same throughout, will be men 
tioned later. 

Consider a brass-mill mold, either 
round or rectangular in cross section, 
standing several feet high. It may require 
a minute or two for pouring. Before it is 
full, the metal in the lower end has partly 
solidified and contracted somewhat away 
from the mold. The initial shell of solidifica 
tion follows not far behind the top of the 
column, even if the central portion is 
liquid for a long way down. There is, 
then, a space between metal and mold 
that is filled with zinc vapor. The vapor 
has no exit, because the liquid at the top 
of the rising column is not only in contact 
with the mold but keeps up a fresh supply 
of zinc vapor, which has, perhaps, already 
begun to condense in the lower portion. 
The cycle is short. Within a minute or 
two after the mold is full the bar has 
cooled enough to condense all the zinc. 
Some of it may condense on the mold; 
it might be expected that most of it would 
condense there, but the mold at that time 
is well coated with a velvety carbon residue 
from the oil dressing, overlain by fluffy 
zinc oxide that commenced covering 
the inside of the mold as soon as pouring 
began. The clean skin of the freshly cast 
brass offers a preferred surface for con 
densation. Reabsorption is quick and 
inevitable. 

A simple experiment indicates that a 
surprisingly large quantity of zinc vapor 
is trapped in the space between mold and 



DANIEL R. HULL 



99 



metal. This was suggested by the observa 
tion that beads of zinc were frequently 
found adhering to the lower end of billets 



setting it upright, to throw in some 
"black lead," which was supposed to 
float up with the metal, give a smooth 



such as are described in Table i. A clue casting and facilitate its withdrawal. 




FIGS. 1-4. TUBE CASTINGS. CROSS SECTIONS AT SURFACE. X 75- 
Nominal composition: copper, 67 per cent- zinc, 32.50; lead, 0.50. 
Fig. i. Normal surface. 
Fig. 2. Surface exudations of lead. 
Fig. 3. Surface exudations of lead. 
Fig. 4. Surface exudations of lead, 



to the origin of beads lay in their occurrence 
only on the bottom end of the billet. It 
was customary, after oiling a mold and 



Perhaps it did sobut enough of it 
remained near the bottom to smother 
and envelop the zinc vapor there, and 



IOO 



SOME PRACTICAL OBSERVATIONS ON INVERSE SEGREGATION 



cause it to condense in discreet drops 
that could not be reabsorbed, although 
they might be and often were attached to 
the parent metal through an alloy bond. 
The effect could be facilitated by con 
tinually throwing black lead into the mold 
throughout the pour, which produced 
zinc condensation clear to the top, and 
in enormously greater quantity. Instead 
of occasional beads, when such a billet 
was pulled out of the mold, large areas 
were sheathed with zinc. 

White beads chipped from a billet like 
that described in Table i were, some years 
ago, reported by the laboratory as a zinc- 
rich copper alloy containing no lead. 
Other beads, attached by an alloy bond 
to a billet containing 0.50 per cent Pb, 
were found to consist of 3.70 per cent Cu; 
balance zinc, again without any lead. 

The absence of lead is significant, and 
fits in with the idea of a condensation 
product rather than an exudation. The 
manipulation with black lead is cited to 
indicate that there is enough zinc trapped 
between metal and mold to accomplish 
the surface enrichment observed. 

GAS SEGREGATION 

The absence of lead enrichment in 
brass surfaces that show normal zinc 
enrichment is a common condition, but 
not invariable. Price and Phillips 4 found 
lead enrichment in surface protuberances, 
which seems to the author to be char 
acteristic of brasses that are liable to 
gassing. While it is true, as mentioned by 
Genders, 1 that yellow brasses seldom 
give evidence of gas, they are not immune. 
The richer copper-zinc alloys may be 
made to spew by simple overheating and 
pouring into oiled molds: 70-30 may shrink 
indifferently and show hydrogen porosity 
if treated with wet charcoal. Even 65-35 
will fail to shrink, and show hydrogen 
porosity if made, for instance, in a freshly 
lined Ajax Wyatt furnace that has not 
been sufficiently dried. The presence of 



tin or lead will increase the liability of 
brasses to dissolved gas. Admiralty shows 
more evidence of it than straight 70-30, 
and leaded alloys of So per cent Cu or 
more will spew in the mold with moderate 
overheating under charcoal. Exudations 
from all such alloys are rich in tin or lead. 
This is shown in a flat slab of leaded 
rich low brass, 8% by i in. in cross section 
by some 50 in. long, which was covered 
with little spikes protruding from the 
surface. The composition of surface and 
center is given in Table 2. 

TABLE 2. Slab of Leaded Rich Low Brass 



Sample 


Copper, 
Per 
Cent 


Zinc, 
Per 
Cent, 
(Diff.) 


Lead, 
Per 
Cent 


0.03 in. planed from surface. 


83.01 

84.. 43 


13.86 
13 . 63 


3-13 

i 94 


o 10 in next layer 


84. 93 


13- 12 


1.95 




85 . 10 


13 . 25 


i . 55 


Center . 


85. 55 


12.73 


I. 72 











The spikes on the surface are evidence 
of gas evolution, and the zinc enrichment 
at the surface is greater than is ever found 
in normal brass. 

Again, brass tube castings (67 per cent 
Cu, 0.50 per cent Pb, balance zinc) are 
particularly liable to exudation of lead 
in minute protuberances. Some of these 
are shown in Figs, i to 4. Fig. i is a normal 
surface; Figs. 2, 3 and 4 are from the same 
tube, showing the surface in cross section, 
through a protuberance. The outside 
layer of exuded metal is nearly all lead. 
If the alloy is poured into a mold without 
any oil dressing, this type of protuberance 
is absent, and no lead enrichment can be 
found in the surface. The condition is 
modified by reducing the quantity of oil. 
It is not at all stretching the point to 
say that mold oil may contribute to the 
gassing of metal, or accentuate the effect 
of any gas that it may contain. The 
conditions under which this phenomenon 
occurs and fails to occur are cited as 
evidence that gas is the cause. 



DANIEL R. HULL 



101 



The effect of gas in a leaded alloy was 
illustrated impressively, if rather roughly, 
some years ago, when an occasion arose 
for preparing an alloy of 70 per cent Cu, 
29 per cent Pb, i per cent Ni. It was 
melted in a Northrup high-frequency 
furnace and end-poured into bars 4> by 
434 by 60 in. The composition, when 
liquid, had an avid capacity for reducing 
gas, much of which was expelled in freezing. 
When melted under charcoal and poured 
into an oiled mold, the surface of every 
bar was sheathed in lead; but if the copper 
and nickel were first melted and well 
oxidized, the lead then added and the 
whole deoxidized with phosphorus and 
poured into a mold coated only with dry 
carbon, there was never any lead on the 
surface. It was evenly, if coarsely, dis 
tributed through the bar. There is no 
doubt that gas was the only considerable 
factor in this instance. It is cited to show 
how gas may work when a liquid remnant 
remains in an already solidified shell. 
If gas is not the sole cause of lead segrega 
tion at the surface, the author has not 
observed any evidence to the contrary. 

TIN SWEAT 

The tin bronzes are so inclined to inverse 
segregation that a great deal of investiga 
tion and discussion centers about them. 
The simplest case of inversion is the 
" spewed" ingot that does not even shrink 
in the mold. This is so generally admitted 
to be caused by a saturation of reducing 
gas that it is here taken for granted. It is 
worth noting, though, that a tin-rich 
liquid of low melting point is visibly 
forced out from the interior of the ingot. 
There must be a concentration of it there, 
which is "normal" segregation. At the 
start, then, the question is not whether 
" normal" segregation occurs, but what 
causes its inversion. The spewed ingot 
covers rather a large range of gas content, 
from the one that rises like a biscuit 
to the one that shrinks almost to com 



pletion, only to spew a few beads at the 
bottom of the shrink hole. The difference 
is of degree only. In any case, the casting 
is honeycombed with gas cavities and 
useless for commercial work. 



Q 




1b 

<D 
* 


Nominal composition 
Copper 92.20 
Phosphorus 0.30 
Tin ISO 


ant, per cent by 

-J C 


-i 


p 


Tin cont 
01 
















1 











\< Cross section - 2 " ->| 

FIG. 5. AVERAGE TIN CONTENT OP CON 
CENTRIC LAYERS TURNED PROM A BILLET OF 
2%-TKCE. DIAMETER. CAST IN AN OILED 
MOLD. 

Nominal composition: copper, 92.20 per 
cent; tin, 7.50; phosphorus, 0.30. 

The next simplest manifestation of 
inversion is ordinary tin sweat on the 
surface of a chill casting. This also is 
widely admitted to be caused by internal 
gas pressure, even by those who distinguish 
between it and a more moderate, less 
obvious tin enrichment in the surface. 
The evidence favorable to such a view is 
impressive. Suppose a casting is taken 
from a chill mold after it is apparently 
solid. It may have cooled until red is 
barely visible, when the sweat suddenly 
appears in such a ghostlike manner that 
one can only say: "A moment ago it 
was not there, and now it is there." 
Apparently the bar that seemed to be 
entirely solid was really far from it. But, 
again, the liquid metal was certainly in the 
interior, where it should be in "normal" 
freezing; and to be liquid at the tempera 
ture at which it appeared on the surface 
it would have to contain upward of 20 per 



102 



SOME PRACTICAL OBSERVATIONS ON INVERSE SEGREGATION 



cent tin. By whatever process the tin-rich 
tag end reached the surface, it must have 
penetrated and passed through the already 
"solid" walls. It does not too seriously 




FIG. 6. CROSS SECTION AT SURFACE, CAST 
BILLET OF 2-INCH DIAMETER, SHOWING LAYER 
OF 20 PER CENT TIN. X 75. 

Nominal composition: copper, 92.20 per 
cent; tin, 7.50; phosphorus, 0.30. 

strain one s credulity to imagine that, 
when freezing approached completion, 
gas thrown out of solution (either through 
concentration or reaction) developed pres 
sure enough to expel this liquid residue 
through the side walls, appropriating 
the space to itself. The characteristic 
unsoundness at the center of such castings 
has generally been assigned to the wide 
freezing range that results in interdendritic 
shrink cavities, cut off from access to any 
source of "feed/ but the freezing range 
might well involve a secondary implication 
that of holding gas in solution until 
all avenues of escape have been closed. 
It is thrown out of solution only to be 
trapped in the center of the casting, under 
substantial pressure. 

There is the view of Price and Phillips, 4 
that the liquid interior is pressed to the 
surface by the contracting shell. This would 
increase the pressure on the entrapped gas, 



and the solid shell of the casting (commer 
cial rolling-mill slabs) would appear strong 
enough to sustain it. But Genders further 
points out that the shell contraction 
ought to close the interior voids. This 
seems reasonable and, since bronze with 
heavy tin sweat is particularly unsound 
at the center, the function of the gas 
seems probable, 

TABLE 3. Composition of Layers from 
Tin Billet 



Layer 


Thickness, 
In. 


Tin, 
Per Cent 


i , surface 


Me 


2"\ 


2, next 


X* 


7-55 


3, next. 


H 


6,92 


4, next. ... 


H 


6.98 


5, next. . 


H 


6.89 


6, next. . . . 


H 


6.95 


7, next. 
8, next. 


H 

H 


6.88 
6.93 


9, center 




6.98 



It has often been pointed out that the 
degree of inverse segregation depends on 
the degree of gassing. This is true in 
commercial work. In a bronze of 8 per 
cent tin, for instance, the surface may 
readily contain upward of 20 per cent tin. 
Table 3 gives the composition of layers 
turned from a billet of 2 -in. diameter by 
6o-in. length. Fig. 5 shows the condi 
tion graphically. This does not tell the 
whole story. Thin scrapings from the 
surface were found to contain 20.34 per 
cent tin. 

Fig. 6 is in cross section of this billet, 
showing the sharp division between the 
rich surface and the metal immediately 
under it, which does not greatly vary from 
there to the center. The rich layer is only 
about 0.012 in. thick, and is easily missed 
in sampling. As has been noted, ordinary 
tin sweat would necessarily contain 20 per 
cent to be liquid at the temperature 
observed. While this is typical of bronze 
melted and poured under reducing condi 
tions, the rich surface layer does not 
occur after the use of an oxidizing atmos- 



DANIEL R. HULL 



103 



phere, which is in favor of the gas theory. 
This has been noted by several others, and 
often confirmed. In a comparatively gas- 
free casting of rolling-mill size, the intensely 
rich skin layer is absent, and the surface is 
only moderately high in tin (Figs. 7 and 8). 

That inversion can never be prevented 
entirely by this means is not an argument 
against the gas theory. Commercially, 
there is certainly no gas-free tin bronze. 
It may be possible to destroy the reducing 
gases by oxidation; but up to the point 
where tin oxide appears in the melt, such 
destruction is only relative. If oxidation is 
carried over into tin oxide the melt 
becomes viscous, and strongly reactive with 
any reducing substance available. The 
few experiments that have been made 
with such metal have shown evidence of 
gas, either by spewing or throwing " spikes" 
from the surface. This is believed to have 
resulted from reaction with carbon in the 
mold, but has not been followed up. In 
any case, such metal would have to be 
deoxidized to be commercially useful, 
which would render it again liable to 
absorption of reducing gas. So, although 
there is no gas-free bronze in everyday 
work, it is significant that inverse segrega 
tion is only slight when gassing is slight, 
and is invariably intense when conditions 
are favorable to gas. 

It has been pointed out by Rosenhain 5 
and others that inverse segregation is 
present in a slush casting and this has 
been cited as evidence against the gas 
theory. While it might be construed as 
evidence against shell contraction, it is 
not incompatible with the gas theory, 
because, while the full effect of gas is not 
exerted until the last metal has frozen, 
gas is thrown out of solution, and begins 
to be effective with the first chill in the 
mold. With the freezing of the first shell, 
gas pressure is immediately established 
within that shell, and some of the selective 
high solute residue resulting from that 
first solidification is pushed to the outside. 



If the pressure available for pushing it out 
there is slight, the resistance to its course 
is also slight. 
The quick start of inverse segregation, 



Nominal composition 
Copper 92. 20 

Phosphorus 0.30 
Tin \ 7.50 



Dry mold 

Top end of 

casting 



Cross section -? I /Q - 
FIG. 7. 





fc 
n. 

t!7 



Nominal composition 
Copper 92.20 
Phosphorus 0.30 
Tin | 7. SO 



Dry mold 

Middle of 

casting 



k Cross section-Z /g" -->! 

FIG. 8. 

FIGS. 7 AND 8. AVERAGE COMPOSITION OP 
CONCENTRIC LAYERS TURNED PROM A BILLET 
OF 2^-INCH DIAMETER, CAST IN MOLD COATED 
WITH DRY CARBON. 

Nominal composition: copper, 92,2 per cent; 
phosphorus, 0.3; tin, 7.5. 

and its progression through the freezing 
period can be traced in the shrink end of a 
vertical casting. This has been done with a 
2>-m. dia. rod, about 60 in. long, cast 
in an iron chill mold, dressed with oil, 
the alloy being 8 per cent Sn-Cu, 0.3 per 
cent P. When the mold was full the rod 



104 



SOME PRACTICAL OBSERVATIONS ON INVERSE SEGREGATION 



was allowed to shrink naturally, without 
feeding. The shrink hole formed a cone- 
shaped hollow, about 6 in. deep. Con 
centric layers were turned off at a point 
near the top, and again about a foot from 
the top, where solidification had been 
completed in a normal manner. The layers 
were analyzed, with results as shown in 
Table 4- 

TABLE 4. Analysis of Concentric Layers 

ROD 2j^-iN. DIAMETER; COMPOSITION, 8 PER 

CENT TIN, 0.3 PER CENT PHOSPHORUS, 

REMAINDER, COPPER 





L 


ayer 


10 In. be 
low Top, 
Per Cent 


2 In. from 
Top, Per Cent 


Sn 


P 


Sn 


P 


Skin 
Next 
Next 
Next 
Next 
Next 
Next 
Cent 


layer o . 


0. 
0. 
0. 
0. 

o. 


oio in 


9.09 
7-79 
7.28 
7-13 
6.95 
6.94 
7.12 
6.95 


0.40 
0.29 


8.14 
7-72 
7-13 

Hollow 


0-34 


10 In .... 


10 in 


20 in . . . 


20 in . . 




20 in 









This indicates that inverse segregation 
had begun, and had proceeded to a sub 
stantial degree in the short time during 
which the mold was full, before the shrink 
went down below the point where the 
top samples were taken, but considerably 
more surface enrichment took place a 
little further down, where it could go to 
completion. In an experiment like the 
above, little spikes or beads can sometimes 
be found protruding from the surface, 
within a fraction of an inch from the top. 
By watching closely, one may see them 
thrown out. They appear after the shell 
has formed, just as the shrink hole is 
starting to sink. The interval is only a 
few seconds, but the suddenness and energy 
with which they appear are strongly sugges 
tive of gas. 

If it is admitted that an intense surface 
concentration of tin can be caused by 
gas pressure, it seems probable that a 
more moderate segregation might be due 
to the same cause. 



CONCLUSIONS 

1. In brass chill castings, the slight 
surface enrichment in zinc is attributed 
to condensation and reabsorption of zinc 
vapor. 

2. In leaded brass, the occurrence of 
lead at the surface is attributed to gas 
pressure. 

3. In tin bronze, inverse segregation of 
tin is attributed to gas pressure. 

ACKNOWLEDGMENTS 

Acknowledgments are due to Messrs. 
H. L. Silliman and J. C. Rowell, of The 
American Brass Company s technical staff, 
for collection and preparation of data, 
and to Dr. Cyril Stanley Smith for con 
structive support and criticism. 

REFERENCES 

1. R. Genders: The Mechanism of Inverse Segrega 

tion in Alloys. Jnl. Inst. Metals (1927) 37, 241. 

2. A. Phillips and R. M. Brick: Segregation in Single 

Crystals of Solid Solution Alloys. Trans. 
A.I.M.E. (1937) 124, 313. 

3. N. B. Vaughan: Inverse Segregation, a Review. 

Jnl. Inst. Metals (1937) 67, 35. 

4. W. B. Price and A. J. Phillips: Exudations on Brass 

and Bronze. Trans. A.I.M.E. (1927) 27, 80. 

5. S. L. Archbutt: Discussion of R. Genders, loc. cit. 

DISCUSSION 

(J. T. Norton presiding) 

W. B. PRICE,* Waterbury, Conn. I con 
gratulate the author on a fine, practical paper. 
I cannot entirely agree with his conclusions that 
inverse segregation is caused entirely by gas 
pressure. Under certain circumstances, it may 
be caused by contraction of the shell or by a 
combination of shell contraction and gas 
pressure. 

G. EDMUNDS,! Palmerton, Pa. Has anyone 
made vacuum casts of these alloys that are 
susceptible to inverse segregation not only 
in a vacuum but of thoroughly degasified 
alloys? 

J. X. KEMP,! Waterbury, Conn. I join 
Mr. Price in welcoming Mr, Hull s paper as 
one of the practical type and that appeals to 

* Chief Chemist and Metallurgist, Scovill Manu 
facturing Co. 

t Research Division, New Jersey Zinc Co. 

j Metallurgical Engineer, American Brass Co. 



DISCUSSION 



men who have arrived through the refinery 
and the mill. We have seen phenomena that 
have been described as inverse segregation 
in the present paper; we have seen them occur 
many times and in many places, and when they 
are brought before us as simply and as beauti 
fully as has been done by Mr. Hull we think 
back to the many cases in which we have seen 
evidence of internal precipitation of gas as the 
cause of the appearance of molten metal 
on the outside of partly solidified castings. 

This phenomenon is not confined to alloys 
whose melting temperatures are largely affected 
by the alloy addition and that have a wide 
liquidus-solidus temperature spread. We have 
seen the same thing in almost pure metals, 
metals whose structure is not unlike that of the 
casting shown in Fig. 6. 

At one time, I had the task of knocking off 
the sink-heads from malleable nickel and 
Monel ingots before they were entirely frozen, 
and occasionally, if I knocked them off a little 
too quickly, before they were completely 
frozen, I would bleed some of the molten metal 
from the head and then find that the space 
within the head was filled with a network of 
dendritic crystals, crystals that might be 
1^2 to 2 in. long. Such an observation shows 
that as the time of final solidification approaches 
there is a mesh of solid crystals in a bath of 
liquid metal. It shows readily the conditions 
that must exist at the moment of gaseous 
precipitation and the development of a pressure 
thereby on the liquid within the casting. It also 
indicates the probability of the simultaneous 
existence of interdendritic channels in the outer 
shell of the castings through which the inner 
liquid can flow. Such channels may still be 
filled with liquid at the critical moment or with 
metal still soft enough to yield to hot liquid 
penetration. 

There are many examples of this behavior. 
The common development of snakes on blister 
copper, an impure copper to be sure, of 96 per 
cent or so, is one. There may be a eutectic in 
blister from which the liquid is ejected; perhaps 
there is, but the general appearance of a 
freezing cake suggests rather that the snakes 
come from the liquid reservoir within. There is 
a time relation too between the appearance 
of ejected liquid copper and the formation of 
the dark area on the crust that marks the 
" blister." Then there are the little spurts of 



flame, burning gas that escapes through the 
crust to the atmosphere, to say nothing of 
the lift of the crust itself. 

I have seen spewing on nickel ingots running 
99.6 per cent Ni. I have seen it on nickel shot 
analyzing 99.8 plus, poured in water. Some 
grades of nickel must be "quieted" in the 
furnace before pouring or the shot will not sink 
in the water tank. Under some conditions there 
is such a volume of gas released from the shot 
as the drops of metal solidify under water that 
it burns at the water surface as the columns of 
bubbles emerge. 

I have seen flames on top of many a solidify 
ing nickel ingot. Such ingots often can be 
broken open with a sledge. When they are 
broken open, particularly those that have cast 
badly and have swelled and run over the mold, 
they exhibit ample interdendritic porosity. 
Metal properly conditioned in the furnace lies 
quietly in the mold, solidifies without visible 
discharge and is sound and hard to break when 
cold. Those that snake badly will be as full of 
holes as a sponge. 

Mr. Hull is right in his belief that it is the 
development of pressure by internal gas 
precipitation from the final liquid phase of a 
casting that is the cause of the appearance of 
liquid on the solid surfaces of cast bars. The 
evidence that he presents, the rich alloy of 
the exudate, is convincing. 

R. M. BRICK,* New Haven, Conn. I have 
always felt that inverse segregation was at 
least one problem we could teach students with 
some assurance, on the basis that the dendrites 
forming in the liquid are enriched in the higher- 
melting-point element. The liquid adjacent to 
those dendrites is enriched in the lower-melting 
element (or elements), and as solidification 
proceeds interdendritic contraction occurs; that 
is, two dendrites contract away from one 
another and open a channel between them, a 
low-pressure area, into which the enriched 
liquid flows to the surface. That ties up very 
nicely with the physical conditions and our 
phase diagrams. 

There is no doubt that the process is expe 
dited or accentuated by internal gas pressure, 
or in some cases by contraction of the shell; 
but I do not believe that it can be urged suc 
cessfully that gas pressure is predominant in all 

* Hammond Laboratory, Yale University. 



io6 



SOME PRACTICAL OBSERVATIONS ON INVERSE SEGREGATION 



cases or is the ultimate cause, since we have 
found inverse segregation in single crystals of 
solid solution alloys where solidification has 
taken place in one direction only, the tempera 
ture gradient extended in one direction only, 
the single crystal froze from the bottom up and 
the surface was always kept liquid. Yet, the 
bottom of the single crystal was enriched in the 
lower-melting-point element. This seems to be 
pretty strong evidence against accepting the 
gas-pressure theory or contraction pressure as 
the predominant or the fundamental cause. 

C. S. SMITH,* Waterbury, Conn The role 
of zinc vapor in producing local differences of 
composition is a new and most interesting 
observation. I suspect that under some con 
ditions an effect greater than the diffusion of 
zinc in and out of the solid casting is the con 
densation on the mold walls of zinc vapor 
evolved by the stream of liquid brass, the zinc 
being subsequently re-absorbed in the quickly 
chilled outer layers of liquid as the mold fills. 
Diffusion from or into the solid alloy could, of 
necessity, affect only a very thin layer. A case 
where this undoubtedly is involved is gamma 
brass. Castings of this alloy invariably have a 
uniform thin layer (0.005 i n -) f a ductile 
material on the surface, presumably beta. I 
believe this results from zinc vaporization at 
the surface of the solid material, the vapor 
condensing on the mold rather than on other 
parts of the casting. It occurs to a greater 
extent in a dry dressed mold than in oiled 
molds, and is absent on the parts of the casting 
that are exposed to air and covered with an 
impermeable, oxide coat. 

A combination of many theories is necessary 
to account for the observed facts of segregation. 
The gas-pressure theory has generally found 
favor among men who have observed large- 
scale commercial operations, while laboratory 
workers tend to favor more involved theories. 
The difference, I think 3 is largely a result of 

* Research Metallurgist, American Brass Co. 



lack of agreement as to what is meant by 
inverse segregation. Surface exudations like 
those shown in Figs. 2 to 4 represent only one 
type. Anyone who sees metal squirt out of the 
surface of a large billet cooling on the casting- 
shop floor will quickly be convinced that gas 
pressure is in large measure responsible for 
these exudations and for the related effects of 
shallow surface enrichment resulting from the 
spreading of exuded low-melting-point ma 
terial. This exudation type of segregation is 
quite different in cause and nature from the 
more gradual change of composition from 
center to outside that is found in substantially 
gas-free melts, and requires a more elaborate 
explanation. 

D. R. HULL (author s reply). In preparing 
my remarks, I was tremendously impressed 
with one thought all through it what a paper 
a man could write if he knew all about this 
subject. 

For a long time I subscribed to the shell- 
contraction theory, but there are some things 
about it that are difficult to explain. For in 
stance, if the liquid is forced to the surface by 
shell contraction, why, as has been pointed out 
before by others, is the inside so unsound? The 
central portion of an ingot that has shown a 
great deal of inverse segregation is notoriously 
unsound. 

Mr. Edmunds asked if we had any results 
from vacuum melting. We have not. I still 
have a feeling that when somebody makes a 
really effective vacuum melt of tin bronze, the 
center will be tin-rich, as it ought to be. 

The interdendritic theory that Dr. Brick has 
spoken of, I can accept up to a certain point. 
I can readily see how the interdendritic view 
might account for tin enrichment near the 
surface, but it is difficult to visualize how that 
could push the tin-rich alloy clear out onto the 
surface in the form of a sweat, or leave the 
center poorer in tin than the surface. 



The Self-diffusion of Silver 



BY WILLIAM A. JOHNSON* 

(New York Meeting, February 1941) 



THE fundamental role of diffusion in 
many reactions occurring in solid metals 
has long been recognized, and there have 
been careful measurements of rates of 
diffusion in numerous alloy systems; but 
our understanding of the atomic mechanism 
by which diffusion occurs is scarcely in an 
advanced state. 1 " 3 It is doubtful that this 
mechanism can ever be directly ascer 
tained; rather, all imaginable mechanisms 
must be examined and the one that is in 
best accord with such data as can be 
obtained from suitably chosen experiments 
should be selected. Because of its relative 
simplicity, diffusion in a system containing 
only one kind of atomic species com 
monly called self-diffusion is of consider 
able theoretical interest. It may be 
expected that diffusion theory will not be 
advanced appreciably until a wide variety 
of diffusion experiments of which diffu 
sion in a chemical concentration gradient 
and self-diffusion are two important 
types are carefully performed and inter- 
compared. Aside from the purely theoretical 
interest in self-diffusion measurements, 
such studies will be of considerable en 
gineering importance. The rate of any 
process in a pure metal requiring the move 
ment of atoms, of which recrystallization 
and creep are two familiar examples, will 
be understood more readily with the aid 
of self-diffusion data.f 

Manuscript received at the office of the Institute 
Nov. 25, 1940. Issued as T.P. 1272 in METALS 
TECHNOLOGY, January 1941- , , 

* Carnegie Institute of Technology, Pittsburgh, 
Pa. Formerly Westinghouse Research Fellow, West- 
inghouse Research Laboratories, East Pittsburgh, Pa. 

1 References are at the end of paper. 

f The determination of the rate of self-diffusion of 
silver is one phase of an investigation of diffusion in 
the silver-gold system by the usual chemical methods 
in a chemical concentration gradient, and by radio 
activity methods in homogeneous alloys. 



Self-diffusion implies the diffusion of 
something into its exact counterpart; the 
detection of such a process is patently 
impossible, and it is necessary to substitute 
for it the diffusion of one isotope (in this 
case radioactive) into another isotope (in 
this case stable) of the same chemical 
element. There is, however, no reason to 
suppose that a radioactive atom will not 
diffuse at the same rate as its stable isotope, 
for all of the isotopes of the same element 
have exactly the same chemical properties; 
neither the difference in nuclear energy 
nor the difference in mass would be ex 
pected to exercise an important influence 
in this phenomenon. In this sense, self- 
diffusion may be thought of as the diffusion 
of radioactive atoms down a radioactivity 
concentration gradient, just as usual diffu 
sion is the diffusion of one kind of atom 
down its chemical concentration gradient. 
The first measurements of self-diffusion 
were made by von Hevesy 4 " 6 using the 
radioactive lead isotope thorium B in 
ordinary lead. The recent development of 
means for producing artificial radioactive 
elements has made possible similar meas 
urements on gold, 7 - 8 copper, 2 - 9 and zinc. 10 - 11 
Many of the data that have been obtained 
are rather poor, and no adequate reason 
has been given for their poorness; it ap 
pears, however, that the unsatisfactory 
results are characteristic of the particular 
experimental procedure employed, and not 
of self-diffusion measurements in them 
selves. The experimental technique that 
has yielded the poorer results is the deposi 
tion of a thin layer of radioactive atoms 
on the surface of a specimen and the deter 
mination of the rate of inward diffusion 



107 



108 



THE SELF-DIFFUSION OF SILVER 



by the decrease of surface radioactivity; 
this procedure has been used in the study 
of lead, 4 5 - 6 copper, 2 and gold. 8 A more 
satisfactory, if more tedious, method is to 



silver chloride. Using this solution with 
standard plating technique, a layer of 
radioactive silver about 0.0004 cm. thick 
was plated on one face of a silver disk 



L 



\ 



FIG. i. CONCENTRATION-PENETRATION CURVE FOR DIFFUSION OF RADIOACTIVE SILVER IN 

ORDINARY SILVER AFTER 4.78 DAYS AT 876.OC. 



section the specimen after diffusion and 
determine a concentration-penetration 
curve, a procedure quite analogous to 
that usually employed in alloy diffusion; 
measurements by this principle have been 
made on copper, 9 gold, 7 and zinc, 10 - 11 
and, as will be described in the present 
paper, on silver. 

EXPERIMENTAL PROCEDURE 

Radioactive silver, initially approxi 
mately equal parts of Ag 105 (half-life 45 
days) and Ag 106 (half-life 8.2 days) was 
prepared by proton bombardment of 
palladium in a cyclotron. 12 The radioactive 
silver was chemically separated from the 
palladium as silver chloride, with 50 milli 
grams of ordinary silver added as a carrier; 
the concentration of radioactive silver in 
the precipitate was very low, of the order 
of magnitude io~ 12 , but this concentration 
does not enter into the subsequent calcula 
tions. A standard silver cyanide electro 
plating solution was prepared from the 



% in. in diameter and Y in. thick. The 
silver disk, in which the grain size was 
about one grain per square millimeter, 
was prepared from Baker chemically pure 
silver; analysis showed the silver to be of 
high purity, except for 0.04 per cent Cu. 
A second similar silver disk was welded to 
the active surface by heating the disks 
under pressure for 30 min. at 75oC. in a 
vacuum furnace. This yielded a composite 
specimen consisting of a thin plane layer 
of radioactive silver bounded on both sides 
by effectively semi-infinite layers of ordi 
nary silver. Diffusion was thus one-dimen 
sional, but occurred in two directions. The 
specimen was then sealed off in a highly 
evacuated quartz tube and placed in a 
potentiometrically controlled electric fur 
nace for the diffusion anneal. After the 
diffusion anneal, layers of thickness from 
0.002 to 0.005 in-? parallel to the original 
active layer, were machined off in a lathe. 
These layers were dissolved in nitric acid 
and the radioactivity of each layer was 
measured by a liquid glass countertube 



WILLIAM A. JOHNSON 



109 



connected to a scale-of-four Geiger-Muller of the diffusion equation 
counting circuit. ^ ^ 

As may be inferred from the technique 
employed, no large experimental errors 



-r- = D 



X 



FIG. 2. REPLOT OF DATA FROM FIG. i. 

Log radioactivity as ordinate and distance from interface squared as abscissa; open circles 
from left side of Fig. i, closed circles from right side. 



are to be expected. The temperature 
measurements are believed to be correct 
within iC.; an error of iC. in the tem 
perature produces an error of about 2 per 
cent in the calculated diffusion coefficient. 
Errors in the measurement of distance 
should produce an error not greater than 
2 per cent. The statistical error in the 
measurement of radioactivity ranged from 
i to 2 per cent, and should introduce a 
much smaller error in the calculated diffu 
sion coefficient. The amount of diffusion 
that occurred during the welding operation 
is negligible; the penetration is determined 
by the product of time and diffusion coeffi 
cient, and this product for the welding 
operation was much less than i per cent 
of the value for the diffusion anneal. The 
several experimental errors are independent 
of one another, and it is probable that the 
calculated diffusion coefficients are accu 
rate to 3 or 4 per cent. 

EXPERIMENTAL RESULTS 

If the initial layer of radioactive silver 
be considered infinitely thin, the solution 



appropriate to the boundary conditions 
employed experimentally is 13 



[il 



where c is the concentration of radioactive 

atoms, 
x is the distance from the original 

interface, 

D is the diffusion coefficient, 
t is the time, 

C Q a constant, is the product of the 
initial concentration of radioactive 
silver times the thickness of the 
layer. 

The error introduced in neglecting the 
finite thickness of the layer of radioactive 
silver is less than i per cent, because of the 
large penetration obtained. 

Fig. i, which is characteristic of all the 
experimental results, is a concentration- 
penetration curve obtained from a specimen 
heated at 876.oC. for 4.78 days. The open 
circles are experimental points and the 
smooth curve is calculated from equation i, 



no 



THE SELF-DIFFUSION OF SILVER 



using the value of D that yields the best fit 
between the calculated curve and the data. 
The good agreement between the data and 
the calculated curve indicates the applica- 



1000 

FIG. 3. VARIATION or SELF-DIFFUSION 

COEFFICIENT OF SILVER WITH TEMPERATURE, 
LOGARITHM OF D VERSUS RECIPROCAL OF 
ABSOLUTE TEMPERATURE. 

bility of simple diffusion theory, and the 
fact that the two sides of the curve are 
symmetrical about the original interface 
shows that the electroplated and welded 
interfaces are equally good. There is no 
evidence supporting the belief that a 
fraction of the radioactive atoms remain 
" stuck" at the original interface, as has 
been reported for the self-diffusion of 
zinc. 10 - 11 

The type of curve shown in Fig. i is not 
convenient for calculating the diffusion 
coefficient. By taking logarithms of equa 
tion i, the following relation is obtained: 

log c- -0.4343 + log 



= 

which may, for the present purpose, be 
more conveniently written 

x 2 
log c = -0.1086 + log k [2] 



If the logarithm of the radioactivity is 
plotted as a function of the square of the 
distance from the interface, the data will 
be expected to lie along a straight line, and 
this expectation is accurately fulfilled, 
Fig. 2. The slope of this line, according to 
equation 2, is 

. 0.1086 

slope = ^- 

and since / is known, D is calculated readily. 
The units of D are square centimeters per 
second. The diffusion coefficient calculated 
in th s manner is slightly too small, since 
diffusion distances are shorter at room 
temperature, where measurements were 
made, than at the diffusion temperature. 
The proper correction is readily calculated 
from the coefficient of thermal expansion 
of silver and varies from 2.6 per cent at 
725C. to 3.5 per cent at 95oC. 

The variation of the diffusion coefficient 
with temperature is shown in Fig. 3, in 
which the logarithm of D is plotted against 
the reciprocal of the absolute temperature. 
The four experimentally determined points 
fall nicely along a straight line, whose 
equation is 

-45.950 

D = 0.895^ RT 

DISCUSSION or RESULTS 

It is of interest to note that the activa 
tion energy for the self-diffusion of silver, 
45,900 cal. per mol, agrees well with the 
figure of 45,200 cal per mol given by the 
Dushman-Langmuir equation. 

It is not unreasonable to suppose that 
this activation energy, Q, should bear some 
relation to the energy required to destroy 
the metal lattice, of which the temperature 
of the melting point, r w , and the binding 
energy,* E, are convenient measures. 2 
The available information on the self- 
diffusion of metals, taken from Steigman, 
Shockley and Nix, 2 with the addition of 



* The binding energy E is the heat of siiblimation 
at absolute zero. 



DISCUSSION 



III 



the present data for silver and recent data 
for zinc, 10 - 11 are given in Table i. Q is in 
nearly constant relationship to T m and E 
in the cubic metals copper, gold, lead and 
silver; the same relationship is borne out 
in the noncubic metals bismuth and zinc 
if the lower of the activation energies 
corresponding to the two principal crystal- 
lographic directions is chosen. As Steigman, 
Shockley and Nix point out, it is not 
possible to say to which of these measures 
of lattice energy Q is more closely related. 

TABLE i. Da ta on Self -diffusion 







Tm, 










Metal 


K Cal. 
per 

Mol 


Deg. 

Melt 
ing 


E 
KCal. 

A/fol 


Q 

Tm 


Q 

E 


Refer 
ence 






Point 










Cu 


57-2 


I3S6 


81.2 


42 


0.70 


2 


Cu 


61.4 


1356 


81.2 


45 


0.76 


9 


Au 


51-0 


1336 


92 o 


38 


0-55 


8 


Au 


62.9 


1336 


92.0 


47 


0.68 


7 


Pb 


28.05 


600 


47-5 


47 


0.59 


14 


Ag 


45-9 


1234 


68.0 


37 


0.68 




Bi _L C 


31-0 


544 


47.8 


57 


0.65 


IS 


Bi || C 


[40 




47-8 


257 


2.92 




Zn |! C 


17.6 


693 


27.4 


25 


0.64 


ii 


Zn_L 


~I7.6 


693 


27-4 


~25 


~o.64 


IO 



a These values are inferred from data obtained on 
polycrystalline zinc. 

No data except those published here 
with have been presented on the rate of 
self-diffusion of silver. The activation 
energy determined in this work appears 
quite reasonable, however, since it bears 
the usual relation to T m and E, since it 
agrees with that calculated from the Dush- 
man-Langmuir equation, and since it is 
somewhat higher than the activation energy 
for the diffusion of other elements in 
silver. 16 

ACKNOWLEDGMENT 

It is a pleasure to thank Dr. J. E. Hill, 
who prepared the radioactive silver in the 
cyclotron of the University of Rochester, 
and the Westinghouse Research Labora 
tories for the granting of a fellowship. 



REFERENCES 

1. R. F. Mehl: Trans. A.I.M.E. (1936), 122. 21. 

2. J. Steigman, W. Shockley and P. C. Nix: Phys. 

Rev. (1939) 56, 13. 

3. R. M. Barren Proc. Phys. Soc. Lond. (1940) 52, 

58. 

4. J. Groh and G. von Hevesy: Ann. Physik. (1920) 

63, 85. 

5. G. von Hevesy and A. Obrutscheva: Nature 

(1925) 115, 674- 

6. G. von Hevesy and W. Seith: Ztsch. Physik (1929) 

56, 790. 

7. A. Sagrubskij: Phys. Ztsch. Sowjetunion (193?) 

12, 118. 

8. H. A. C. McKay: Trans. Faraday Soc. (1938) 34, 

9. B. V. Rollin: Phys. Rev. (i939) 55. 231. 

10. P. H. Miller and H. Day: Phys. Rev. (1940) 57, 

1067. (Abstract.) 

11. F. R. Banks and H. Day: Phys. Rev. (1940) 57, 

1067. (Abstract.) 
.12. T. Enns: P,hys. Rev. (1939) 56, 872. 

13. See, for example: Riemann-Webers: Differential 

Gleichungen der Physik, 2. Braunschweig, 
1927. Vieweg and Sohn. 

14. G. von Hevesy, W. Seith and A. Keil: Ztsch. 

Physik (1932) 79, 197. 

15. W. Seith: Ztsch. Elektrochem. (1933) 39, 538. 

16. W. Seith and E. A. Peretti: Ztsch. Elektrochem. 

(1936) 42, 570. 

DISCUSSION 

(R. M. Brick presiding) 
F. E. CARTER,* Newark, N. J. Fig. 3 shows 
the variation of self-diffusion with temperature. 
Without claiming any particular knowledge on 
this subject, I might have expected a very 
considerable increase in the rate of self- 
diffusion as the melting point is approached. 
I look on this property as somewhat analogous 
to extrusion. The extrudability of a metal 
increases exceedingly rapidly as the tempera 
ture approaches the melting point, indicating 
marked increase in mobility just below that 
point. The author has continued his straight- 
line curve right up to the melting point and I 
would like to ask whether he has considered 
that any marked deviation is possible between 
the highest temperature of his measurements, 
95oC., and the actual melting point of pure 
silver. 

P. H. BRACE,! East Pittsburgh, Pa. What 
is the significance of the term "activation 
energy " ? 

W. A. JOHNSON (author s reply). No direct 
correlation between the phenomena of extruda 
bility and self-diffusion has been discovered, 
and there is at least no obvious reason for 
supposing that a correlation should exist. No 
data for the self-diffusion of silver have been 



* Physical Metallurgist, Baker Platinum Co. 
j" Westinghouse Electric and Manufacturing Co. 



112 



THE SELF-DIFFUSION OF SILVER 



obtained at temperatures very slightly below 
the melting point, but such data are available 
for lead and bismuth, 17 and these obey a 
straight-line relationship between log D and 
i/r, apparently to the melting point, 

A great deal might be said in reply to Mr, 
Brace s question. A very simple explanation of 
the term " activation energy" might be the 
following: A given atom must possess the 
energy e to move from one lattice position to 
another; this energy is its " activation energy." 
The probability that an atom will possess the 
requisite energy for movement varies with 

17 W. Seith: Diffusion in Metallen, 48, 68. Berlin, 
1939. J. Springer. 



temperature according to the well-known 
Boltzmann equation 

p = e -/kT 

so that a measurement of the variation in 
diffusion rate with temperature will furnish a 
means for determining e quantitatively. The 
diffusion coefficient is thought to be propor 
tional to the probability p, the proportionality 
constant involving, perhaps among other fac 
tors, the frequency of vibration of the atoms 
and the interatomic spacing; it cannot be 
exactly evaluated until a satisfactory theory 
of diffusion is devised. 



the corresponding 
equilibrium diagrams are known. 

The 6-versus-temperature curves for 
two alloys PI and P 2 are shown in Fig. i. 



On the Equilibrium Solidification of Solid Solutions 

BY MORRIS COHEN,* MEMBER A.I.M.E., AND WILLIAM P. 
(Cleveland Meeting, October 1940) 

ABSTRACT! 

THIS paper deals with the calculation to-solid transfers if 
of the composition of the infinitesimal 
trace of alloy that transfers from the liquid 
to the solid state at each temperature 
during the equilibrium solidification of a 
binary solid solution, and which thereby 
enables the compositions of the coexisting 
liquid and solid phases to follow along the 
liquidus and solidus lines respectively 
during the cooling. A general expression 
for the composition (5) of the liquid-to- 
solid transfer at each temperature has been 
derived: 



5 = 



t 



- P) + Z s x s (P - x s ] 
Z L (XL -P) + Z S (P -x s ) 



where Z L and Z s are the slopes of the 
liquidus and solidus lines at the given 
temperature, XL and x$ are the composi 
tions of the stable liquid and solid phases 
at the same temperature, and P is the 
composition of the alloy under considera 
tion. The 8 equation is valid for the process 
of equilibrium heating as well as cooling, 
and may be applied to vapor-to-liquid, 
vapor-to-solid, liquid-to-liquid, and solid- 




* Assistant Professor, Department of Metallurgy, 
, Massachusetts Institute of Technology, Cambridge, 
M assachusetts . 

t Professor, Thayer School of Civil Engineering, 
Dartmouth College, Hanover, New Hampshire. 

J Manuscript received at the office of the Institute 
June 27, 1940. The entire manuscript has been sub 
mitted to the _ American Documentation Institute, 
2101 Constitution Avenue, Washington, D. C., for 
issuance in the form of microfilm or photoprint. The 
derivation of the equations and their physical signifi 
cance are discussed in detail in the paper. For entire 
paper, write to the American Documentation Insti 
tute, ordering Document 1440 and remitting thirty- 
five cents for microfilm (read enlarged on machines 
widely available) or one dollar and eighty cents ($1.80) 
for photoprint (readable without optical aid). This 
abstract was issued as T.P. 1256 in METALS TECH 
NOLOGY, December 1940. 



x. P, i, P 2 S, 

WT%B 
FIG. i. GENERAL SOLID SOLUTION DIAGRAM 

SHOWING 5 CURVES FOR ALLOYS PI AND P 2 . 

In each case, the liquid-to-solid transfer 
has the same composition as the stable solid 
phase only at the temperature of the be 
ginning of solidification, and has the same 
composition as the stable liquid phase only 
at the temperature of the completion of 
solidification. At each intermediate tem 
perature, the composition of the liquid-to- 
solid transfer differs by a finite amount 
from those of the two coexisting phases. 
Therefore, during equilibrium cooling, 
the transfers must merge into the solid 
phase from the liquid without forming a 
transition phase; otherwise the condition 
of equilibrium would be violated. At any 
given temperature during solidification, 
the composition of the liquid- to -solid 



113 



EQUILIBRIUM SOLIDIFICATION OF SOLID SOLUTIONS 



transfer is richer in per cent B, the lower 
the per cent B in the alloy as a whole. 

In the general case, the values of Zz,, 
Zs, XL and oc$ which appear in the d 




example, if the liquidus and solidus lines 
are straight over the solidification range 
and intersect on the temperature axis 
at the melting point TA of component A, 
the 5-versus-temperature curve becomes a 
parabola: 

8 = fr. . 



1,0 20 

WT % CARBON 

FIG. 2. PORTION OF IRON-CARBON DIAGRAM 
SHOWING S CURVES FOR ALLOYS CONTAINING 
0.83 PER CENT AND 1.50 PER CENT CARBON. 

equation are determined graphically from 
the equilibrium diagram. However, if the 



where WIL and ms are the constant slopes 
of the liquidus and solidus lines. In such a 
straight-line system, it turns out that the 
weight of component B contained in each 
liquid-to-solid transfer is independent of 
the temperature and the composition of 
the alloy, even though equation 2 shows 
that the percentage of B in the transfer 
depends upon both these factors. Accord 
ingly, variations in the composition of the 
liquid-to-solid transfers must be accom 
plished merely by variations in the weight 
of component A contained in the transfers. 
Calculated 5 curves for five iron-carbon 
alloys are shown in Figs. 2 and 3. The 
curves in Fig. 3 have interesting discon- 




1400 



1.0 



.2 .4 .6 .6 

WT % CARBON 
FIG. 3. PORTION OF IRON-CARBON DIAGRAM SHOWING 5 CURVES FOR ALLOYS CONTAINING 0.04 

PER CENT, 0.14 PER CENT AND 0.28 PER CENT CARBON 

The 8 curves give compositions of solid-to-solid transfers corresponding to conversion of delta 
ferrite into austenite during equilibrium cooling. 



equations of the liquidus and solidus lines 
are known, 5 may be expressed explicitly 
as a function of the temperature. For 



tinuities because of the peritectic reaction 
that interrupts the normal course of 
solidification. 



Measurement of Irreversible Potentials as a Metallurgical 

R esearch Tool 

BY R. H. BROWN,* W. L. FINK,* MEMBER A.I.M.E., AND M. S. HUNTER* 

(Cleveland Meeting, October 1940) 



EARLY workers attempted to study the 
structure of alloys by measurement of 
equilibrium electrode potentials in aqueous 
solutions containing ions of the metals 
from which the alloy was made. 1 The 
method was subject to such limitations 
(e.g., the difficulty in attaining electro 
chemical .equilibrium at low temperatures) 
that it has not been widely used by 
metallurgists. 

Recently, Seltz and co-workers have 
introduced a greatly improved method for 
studying alloys by measurement of equilib 
rium electrode potentials in molten salts, 2 " 4 
The mathematical treatment is the same, 
of course, as for equilibrium potentials of 
amalgams in aqueous solutions. 5 This 
method will probably find extensive use, 
since it makes possible not only the deter 
mination of phase fields of equilibrium 
diagrams but also the determination of 
certain thermodynamic quantities which 
can be used advantageously in metal 
lurgical calculations. However, the method 
presents difficulties. With many metallic 
and alloy electrodes, reversibility is not 
readily obtainable. Also, the difficulty of 
obtaining suitable material from which to 
make the cells is increased as the temper 
ature is increased. Moreover, the necessity 
of carrying out measurements in a vacuum 
requires additional apparatus and intro 
duces considerable difficulty, especially at 
the higher temperatures. 

Manuscript received at the office of the Institute 
Feb. 23, 1940. Issued as T.P. 1234 in METALS TECH 
NOLOGY, October 1940. 

* Aluminum Research Laboratories, New Kensing 
ton, Pa. 

1 References are at the end of the paper. 



Apparently, no consideration has been 
given heretofore to the metallurgical use of 
irreversible electrode potentials that exist 
when a specimen is immersed in a suitable 
aqueous solution at room temperature. 
The apparatus and the method required for 
such measurements are very simple, and if 
the results of such measurements could 
be demonstrated to be reproducib]e they 
might be very valuable in studying the 
structure of alloys. The method would not 
have the scope of that developed by Seltz 
and his co-workers because no thermo 
dynamic quantities could be calculated 
from the data. However, the utility and the 
simplicity of the method might warrant 
extensive use for the determination of 
phase fields and similar applications. 

THEORY or IRREVERSIBLE POTENTIALS 

A brief review of the theory of irrever 
sible potentials will be given before showing 
how to interpret the potential readings in 
terms of alloy structure. It is known that 
heterogeneity of a metal surface in an 
electrolyte will produce points of different 
potentials, with the result that current 
flows between these points and electro 
chemical reactions occur. The points at 
which oxidation reactions are localized are 
known as anodes, and those at which 
reduction reactions are localized are known 
as cathodes. 6 7 - 8 For the benefit of those 
who are not familiar with the terms anode 
and cathode as applied to a primary cell, 
the potential difference between anode and 
cathode of such a cell can be measured when 



n6 



MEASUREMENT OF IRREVERSIBLE POTENTIALS 



the negative terminal of a potentiometer 
is connected to the anode and the positive 
terminal is connected to the cathode. 
The potentials of anodes**and cathodes 



5 
O 




CURRENT - 
FIG. i. SCHEMATIC REPRESENTATION or 

POLARIZATION. 

vary with the current density on them. 
The change of the potential of an anode or 
a cathode resulting from current flow is 
known as polarization. Polarization is 
always in such a direction that the poten 
tials of the anode and cathode approach 
each other as the current through the cell 
increases. Hence, if the cell is short- 
circuited, as are local electrodes on a single 
piece of metal immersed in an electrolyte, 
the potentials of the anode and cathode 
are the same. This potential is the one that 
is measured when a piece of metal, com 
posed of local electrodes, is immersed in an 
electrolyte. 

The relations just described are shown 
diagrammatically in Fig. i ; a is the poten 
tial of the anode and c the potential of the 
cathode when no current is flowing between 
them. When current flows, the potential of 
the anode follows the anodic polarization 
curve a,b,e and the potential of the cathode 
follows the cathodic polarization curve 
c,d,e. The intersection of the curves at 
point e represents the limiting value of the 
current. As a result of this polarization , the 



potential corresponding to point e exists 
over the entire surface of the electrode and 
is, therefore, the potential of the metal 
measured against some reference electrode. 





CURRENT CURRENT CURRENT 

A B C 

FIG. 2. THREE TYPES OP POLARIZATION DIA 
GRAMS (AFTER EVANS). 

Obviously, the measured potential may 
have any value between a and c, depend 
ing upon the slope of the polarization 
curves. 9 " 10 The slopes of these polarization 
curves depend both upon the nature of the 
electrolyte and the area of anode and 
cathode. Other variables being constant, 
the slope of the polarization curve will be 
inversely proportional to the area of the 
local elements. 11 " 12 The possible relations 
are shown schematically in Fig. 2, where 
the polarization curves are represented by 
straight lines for simplicity. 

Fig. 2A shows the relations that obtain 
when the anodic area is very large, or when 
there is little polarization of the anode, 
even with high current densities, and when 
the cathodic areas are polarized sub 
stantially. In this, the case measured volt 
age e closely approximates the potential of 
the anode when no current is flowing. 
Fig. 2B represents the conditions that 
obtain when the anodic area is very small, 
or when there is little polarization of the 
cathode, even at high current densities. 
In this case, the measured potential e 
closely approximates the potential of the 
cathodic areas when no current is flowing. 
Fig. 2C represents the conditions that 
obtain when both cathodic and anodic 
areas are highly polarized; the measured 
potential e will have values between the 
potentials of the anode and cathode. 



R. H. BROWN, W. L. FINK AND M. S. HUNTER 



117 



If the area of the anode phase is small 
and it is desirable to detect its presence by 
potential measurements, the amount of 
anodic polarization should be reduced to a 
minimum. Conversely, if one desires to 
detect the presence of a cathode phase, the 
amount of cathodic polarization should be 
reduced to a minimum. 

DEVELOPMENT OF METHOD 

The choice of the solution in which to 
make the measurement will depend upon 
the type of polarization desired, and also 
upon the open-circuit potentials of the 
local anodes and cathodes. Although the 
known effect of certain ions on polarization 
will help in the selection, a satisfactory solu 
tion can be obtained only by experiment. 

Early in the work on aluminum alloys, 
it was found that neutral chloride solutions 
gave more reproducible potentials than did 
other common electrolytes. However, the 
time required to reach a steady reading 
with solutions of neutral chlorides alone 
was often as long as 24 hr. Also, fluctuations 
in potential frequently occurred, probably 
as a result of variations in oxygen content 
of the solution. It was found that the 
addition of hydrogen peroxide to the 
chloride solution minimized these diffi 
culties. This work led to the choice of a 
solution containing 53 grams of sodium 
chloride and 3 grams of hydrogen peroxide 
per liter. 

In order to ensure reproducible results, 
it was necessary that a standard method of 
surface preparation be developed. Many 
methods were tried. One of the most repro 
ducible surfaces was obtained by machin 
ing, using a very light cut in the final 
operation. Machining was not adopted for 
the standard preparation of surfaces, how 
ever, because the method was expensive 
and time-consuming, and because it was 
difficult to apply to sheet specimens. The 
method that was selected, and which gave 
results that were almost as consistent as 
those of a machined surface, was as follows: 



The specimen was first degreased with 
carbon tetrachloride, then abraded with 
No. 120 emery paper, and finally abraded 
slightly with No. oo steel wool. The steel 



*J -0.76 

-- 74 

2|2 

j "0.72 

g 

i -0.70 

-0.68 




o-l% Cu ALLOY-50% OF TOTAL EXPOSED AREA 
-\% Cu ALLOY-32.5% OF TOTAL EXPOSED AREA 

Cu ALLOY- 14% OF TOTAL EXPOSED AREA. 
*-!% Cu ALLOY- 6% OF TOTAL EXPOSED AREA 

1% Cu ALLOY- 1% OF TOTAL EXPOSED AREA 



04 8 12 16 20 24 28 32 

CURRENT -MILLIAMPERES 

FIG. 3. EFFECT OF POLARIZATION ON POTEN 
TIAL OF DUPLEX ELECTRODES. 
Anodic alloy, 0.97 per cent Cu; cathodic 
alloy, 4.25 per cent Cu. 

wool had sufficient cutting action to remove 
small particles of emery that were em 
bedded in the surface in the previous 
operation. 

Other features of the method were 
standardized to assure reproducible results. 
A o.io normal calomel electrode was used 
as a reference electrode. The potentials 
were measured on a Leeds and Northrup 
type K-2 potentiometer. The temperature 
was held at 25 iC. 

BINARY ALUMINUM-COPPER ALLOYS 

When this method was used, an alumi 
num-copper solid solution high in copper 
was found to be cathodic to an aluminum- 
copper solid solution low in copper. For 
example, a couple formed by connecting an 
aluminum-copper solid solution containing 
i per cent Cu through a sensitive milli- 
ammeter to an aluminum-copper solid 
solution containing 4.25 per cent Cu 
produced a current flowing in the direction 
that demonstrated that the latter solid 
solution was cathodic. However, the cur 
rent flow and the potential of the short- 



n8 



MEASUREMENT OF IRREVERSIBLE POTENTIALS 



circuited couple varied with the ratio of 
exposed electrode areas, as shown in Fig. 3. 
In this figure polarization curves are shown 
for anodic electrodes (i per cent Cu alloy) 



=12 



SOLUTK 


N 53 GRAMS Na 

1 


Cl AND 


1 
3 GRAMS HgQjP 


:R LITER 




H 


i o 










/ 














I 




























) 20 40 60 80 100 0.97% Cu ALL 
30 80 60 40 20 4.25% Cu ALL 



PERCENTAGE EXPOSED AREA 

FIG. 4. POTENTIAL OF ELECTRODES COM 
POSED OF TWO ALUMINUM ALLOYS OF DIFFERENT 
COPPER CONTENTS IN ELECTRICAL CONTACT. 

Anodic alloy, 0.97 per cent Cu; cathodic 
alloy, 4.25 per cent Cu. 

of various areas and a polarization curve 
for a fixed area of the cathodic electrode 
(4.25 per cent copper alloy). The areas of 
the anodic electrodes have been expressed 
as percentages of the total areas of both 
electrodes exposed to the solution. As would 
be expected, the anodic polarization de 
creases as the area of the anode is increased. 
The intersection of the anodic polarization 
curves with the cathode polarization curve 
is the magnitude of current that will be 
obtained when these anodes of various 
areas are short-circuited to the cathode. 
The value of potential at these intersections 
would be expected to correspond to the 
measured potential of two-phase alloys 
composed of various amounts of solid 
solutions containing i per cent and 4.25 
per cent Cu, respectively. The potentials 
at which these curves intersect (e^ e 2 , etc.) 
were plotted against the percentage areas 
of either electrode exposed to the solution, 
and the curve shown in Fig. 4 was obtained. 
This curve shows that as the area of the 
most anodic alloy (i per cent Cu) is 
increased, the potential rises and reaches a 
value practically equal to the potential of 
the most anodic alloy. 



Also, a large specimen of CuAU* coupled 
to any one of a series of aluminum-copper 
solid solutions produced a couple that had 
the same potential as the same aluminum- 
copper solid solution that was not con 
nected to the CuAl 2 cathode. As in the case 
of the couples composed of two aluminum 
solid solutions of different copper contents, 
this demonstrates that in the standard solu 
tion the cathodic polarization predominates. 

Hence the potential of an aluminum- 
copper alloy would be controlled by the 
amount of copper that is in solid solution. 
For example, an aluminum-copper alloy 
containing 4 per cent Cu but having only 
2 per cent Cu in solid solution would have 
the same potential as a 2 per cent Cu alloy 
with the copper entirely in solid solution. 
Also, the potential of an aluminum-copper 
alloy will be independent of the heat- 
treating temperature, providing all the 
copper is in solid solution at all of the heat- 
treating temperatures. Therefore, by means 
of potential measurements in a solution con 
taining 53 grams of sodium chloride and 3 
grams of hydrogen peroxide per liter, it 
should be possible to determine the amount 
of copper in an aluminum-copper solid 
solution. 

A series of aluminum-copper alloys in the 
form of cold-rolled sheet containing from o 
to 7 per cent Cu was heat-treated at five 
different temperatures for 16 hr. and 
quenched in cold water. Assuming that a 
treatment of 16 hr. at these temperatures 
is sufficient to produce equilibrium, and 
that the quenching was of sufficient 
rapidity, the solid solubility at these 
temperatures was deduced. 

The measured potentials of the samples 
heat-treated at each temperature were 
plotted against the copper content, and 
gave a curve with two branches one a 
curved line with a pronounced slope, and 
the other a straight horizontal line. The 
curves for all temperatures were super- 



* The potential of CuAla by itself is 0.53 volts 
on the o.iN calomel scale. 



R, H. BROWN, W. L. FINK AND M. S. HUNTER 



119 



imposed as shown in the lower part of 
Fig. 5.* Since the potential is controlled by 
the copper content of the solid solution, the 
curved line indicates a change in copper 



TEMPERATURE- *C 
S S 8 c 


















^ 


^ 


~^ J 








d\ 


V 




^-POTENTIAL MEASUREMENTS 
o-DIX AND RICHARDSON 

















S ~- 84 

_/ 

I -0.80 

==-0.76 



-0.64 



SOLUTI 

V 


DN 53 G 


RAMS Na 


Cl AND 


3 GRAM 


H&P 


:RUTER 


\ , 














\ 














> 


V 
















^ 










LEG 
-; 


END 
OO C 
100 C 




N 








> * 




v 






A-475 C 
o-500 C 
~525*C 






s 


<^~ 






01234567 

PERCENTAGE COPPER BY WEIGHT 

FIG. 5. POTENTIAL OP BINARY ALUMINUM- 
COPPER ALLOYS AND THE DERIVED SOLID 
SOLUBILITY CURVE. 

content of the solid solution and the 
straight line indicates no further change in 
it; therefore the point of intersection of the 
two lines clearly shows the limit of solid 
solubility. 

The solubility curve as determined in the 
lower part of Fig. 5 is plotted in the upper 
part of the figure in comparison with the 
curve obtained by Dix and Richardson. 13 
The two curves check very well except at 
the two lowest temperatures. From the 
thermodynamics of perfect solutions, it has 
been shown that a plot of the reciprocal of 
the absolute temperature against the 
logarithm of the solid solubility expressed 
in atomic per cent should be a straight line. 
The data on the solid solubility of copper in 
aluminum obtained by potential measure 
ments did fall on a straight line when so 
plotted. 

* On the curved line, only the points obtained at 
S25C. are indicated because the points on the 
curved lines obtained at lower temperatures coincide 
with those obtained at 525C. 



TERNARY ALLOYS 

It should be possible in a like manner to 
determine the limits of solid solubility and 





I 



%c 

FIG. 6. POTENTIAL OF SERIES OF ALLOYS 
ALONG SECTION d-f OF HYPOTHETICAL DIAGRAM 

ABOVE. 

some of the phase fields of ternary alloys. 
In the upper part of Fig. 6 is shown an 
isothermal section of a hypothetical ternary 
diagram. In this section of the diagram, 
element B has practically no solid solu 
bility in A, whereas C has a limited solid 
solubility, indicated by g. If a series of 
alloys along the section a-f behaves in the 
same manner in an electrolyte as the 
aluminum-copper alloys did in the hydrogen 
peroxide-sodium chloride solution, the 
potential curve shown in the lower half of 
Fig. 6 should be obtained. From a to 6, 
the potential will become more cathodic 
because the amount of C in solid solution is 
increasing. The potential will remain con 
stant for alloys from b to c because the 
composition of the solid solution does not 
change. Beyond c and until d is reached, 



I2O 



MEASUREMENT OF IRREVERSIBLE POTENTIALS 



the potential will again become more 
cathodic because the amount of C in solid 
solution is increasing. However, beyond d, 
the potential will not be changed by 



a -0.84 
ri -0,80 
5 -0.76 



< -0.68 



-0.64 



-0.60 



sc 
V. 


LUTION 


53 GR 


AMS Ni 


Cl AND 


3 GRAFi 


IS H 2 2 


PER L 


I 
TER 


\ 










LEG 


END 






1 


^ 


Sy 






*-475 C 
o-500 C 
-525*C 










\ 


*x 


















** 


X, 


-3 


















^v. 



























0123456789 
PERCENTAGE COPPER BY WEIGHT 

FIG. 7. POTENTIAL or SERIES OF ALUMI 
NUM-COPPER ALLOYS CONTAINING 0.35 PER CENT 
IRON. 

further additions of C because the concen 
tration of the solid solution is unchanged. 
Aluminum- copper-iron Alloys. Since 
iron is practically insoluble in aluminum 



ui -0.84 

s 

_j -0.80 
5 -0,76 
.-0.72 
g -0.68 
-0.64 
-n.en 


S( 


(LUTION 


53 GR, 


MS Na 


:i AND 


3 GRAM 


SHA 


3 ER LIT 


:R 


\ 










LEG 


END 










^\ 






A-475C 
o-500 C 
525*C 












X 


N 






























X 


^ 


-a 
















X 


% 























0123456789 
PERCENTAGE COPPER BY WEIGHT 

FIG. 8. POTENTIAL or SERIES OF ALUMI 
NUM-COPPER ALLOYS CONTAINING ONE PER CENT 
IRON. 

and copper has a limited solubility in 
aluminum, two series of aluminum-copper- 
iron alloys were prepared in order to test 
the relations (described in previous section) 
between potentials and structure. In Fig. 7 
are shown the potentials of a series of 
aluminum-copper-iron alloys containing 
0.35 per cent Fe and from o to 8 per cent 
Cu, which had been heat-treated for 16 hr. 
at three different temperatures. As for the 



binary aluminum-copper alloys, the sheet 
was in the hard-rolled temper prior to 
heat- treatment. In Fig. 8 a similar curve is 
shown for a series of aluminum-copper-iron 



<x(AI-Cu) 
Al)+o.(AI-Fe-Cu)+a(AI-Cu) 




AAAAAAA 




Al 123456789 10 

PERCENTAGE IRON BY WEIGHT 

FIG. 9. ALUMINUM CORNER OF ALUMINUM- 

COPPER-IRON CONSTITUTIONAL DIAGRAM AT 



alloys containing i per cent Fe and from o 
to 8 per cent Cu. The shape of these experi 
mental curves is similar to those indicated 




Al 



I 



10 



345678 
PERCENTAGE IRON BY WEIGHT 

FIG. 10. EFFECT OF TEMPERATURE ON 
PHASE BOUNDARIES OF ALUMINUM CORNER OF 
ALUMINUM-COPPER-IRON SYSTEM. 

in Fig. 6. As for the binary aluminum- 
copper alloys, the curved portions of these 
diagrams are similar regardless of the heat- 



R. H. BROWN, W. L. FINK AND M. S. HUNTER 



121 



treating temperature. The two horizontal 
sections are altered by a change in the heat- 
treating temperature. An isothermal sec 
tion at 525C. of the aluminum corner of 



-0.96 
uj -0.92 
rf -0.88 
5^-0.84 
b -0.80 

fe 

- -0.72 

-0.68 



SOLUTIC 


N 53 


iRAMS 


laCI A 


!D 3 OF 


AMS H Z 2 PER 


LITER 




cob 


MERCIA 


. Al + 


i-SSSZn 


t-l.5%0 


u-t-0.45 


IMn-^ 


. - 


, 


- 






















con 


IMERCIA 


. AI + 


4%CiH 


05% M 


+0.5% 


Mn^ 








/ 


^ 


















f 




















I 
r 





















ISO 200 



20 40 60 80 100 120 140 
HOURS AT 150 C 

FIG. n. ISOTHERMAL DECOMPOSITION OF 

SUPERSATURATED SOLID SOLUTIONS. 

the aluminum-copper-iron constitutional 
diagram obtained from Figs. 7 and 8 is 
shown in Fig. 9. This section is consistent 
with the aluminum corner of the diagram 
reported by Bradley and Goldschmidt. 14 

The employment of a triangular diagram 
enables the investigator to check the 
accuracy of his results. For example, a 
check on the location of the boundary 
between the two-phase field of aluminum 
solid solution (Al) and the a(Al-Fe-Cu) 
constituent; and the three-phase field of 
aluminum solid solution, a(Al-Fe-Cu) con 
stituent, and a(Al-Cu) constituent can be 
made in the following manner: A straff ht 
line drawn through the points on this 
boundary for the i per cent Fe alloy and 
the 0.35 per cent Fe alloy should intersect 
the o per cent Fe line at the solid solubility 
limit of copper in aluminum at the temper 
ature in question. Also, a check may be 
made on the other two-phase boundaries 
shown in Fig. 9 in a similar manner. The 
two lines passing through the points on 
these boundaries for the 0.35 per cent Fe 
and the i per cent Fe alloys should intersect 
the o per cent Fe line at the same point. 
This check has been applied to the data 
obtained from Figs. 7 and 8 and is shown in 
Fig. 10. 



DECOMPOSITION OF SOLED SOLUTIONS 

It was found that irreversible potential 
measurements made at 25C. can be used to 



2)2-0. 



-0.65 



SOI 


UTION i 


3 GRAM 


I 1 t 
S NaCI AND 3 GRAMS HjQa PE 


R LITE 


* 








^ 










^ 


<0 


1MERCI 


M Al + 


4%Cu + 


0.5% M 


+ 0.5% 


Mn 


/ 























































20 



60 80 100 120 WO 160 180 
SECONDS DELAY 

FIG. 12. DECOMPOSITION OF SOLED SOLUTION 
CAUSED BY DELAY OF QUENCH. 

gain an insight into the mechanism of 
decomposition of solid solutions. Two 
examples are shown in Fig. n. The poten 
tial of commercial aluminum containing 
4 per cent Cu, 0.5 per cent Mg, and 0.5 per 
cent Mn was found to become more anodic 
the longer it was heated at isoC, This 
would indicate a gradual decrease in the 
copper content of the aluminum solid 
solution at that temperature. On the other 
hand, commercial aluminum containing 
5.5 per cent Zn, 1.5 per cent Cu, and 0.4 per 
cent Mn showed, first, a potential change 
in the cathodic direction during the initial 
period at isoC., which was followed by a 
change in potential in the anodic direction 
during the subsequent stages. It has been 
shown that zinc in solid solution alters the 
potential of aluminum in an anodic direc 
tion, whereas copper in solid solution alters 
the potential in a cathodic direction. 
Because the potential became more ca 
thodic during the initial stages, it was 
indicated that the zinc content of the solid 
solution was decreasing. The subsequent 
shift of the potential in the opposite direc 
tion indicated that the copper content of 
the solid solution was decreasing after pro 
longed heating. These results would indi 
cate that the decomposition of such an 
alloy is at least a two-step phenomenon. 



122 



MEASUREMENT OF IRREVERSIBLE POTENTIALS 



Another example of the decomposition 
of solid solution is shown in Fig. 12. Speci 
mens (4 by i by 0.064 in.) of a commercial 
aluminum alloy that contained 4 per cent 
Cu, 0.5 per cent Mg, and 0.5 per cent Mn 
were heat-treated at 5ioC. in a nitrate 
bath for 15 min., held in air for various 
periods of time and quenched in cold water, 
and their potentials were determined. The 
results show that the longer the specimen 
was held in air prior to quenching in cold 
water, the more anodic the potential of the 
alloy became. This would indicate that the 
longer the delay before quenching, the less 
the amount of copper retained in solid 
solution, or, in other words, the greater the 
decomposition of the solid solution. 

The suitability of this method for deter 
mining concentration of solid solution, or 
the most anodic regions of the solid solu 
tion, suggests a number of metallurgical 
problems in the solution of which irrever 
sible electrode potentials might be very 
useful. Diffusion studies at once suggest 
themselves; in fact, the method has already 
been used to a limited extent for studying 
the extent of diffusion. Segregation, coring, 
rate of solution, identification of alloys, 
determination of porosity in metallic and 
perhaps other types of coatings, all offer 
possibilities for the advantageous use of the 
method described in this paper. 

SUMMARY 

The electrochemical method described 
above is being used advantageously in 
metallurgical investigations of various 
types, such as solid solubility determi 
nations, diffusion studies, age-hardening 
problems, determination of phase fields. 
The basis of the method is the theory of 
irreversible electrode potentials, which has 
been explained and expanded. The equip 
ment and technique required are relatively 
simple, and the measurements can be made 
with considerable rapidity. 

Typical data of this kind have been 
presented on binary and ternary aluminum- 



base alloys, which illustrate the utility 
the method. 

ACKNOWLEDGMENT 

This work was initiated as a result 
preliminary unpublished work by Mr. E. I 
Dix, Jr. and Dr. L. J. Weber. Mr. D. ( 
Sprowls made some of the potenti 
measurements. Dr. H. R. Freche checke 
the identification of the phases in tl 
aluminum-copper-iron system by X-ra 
diffraction measurements. The autho 
gratefully acknowledge the interest an 
suggestions by other members of the sta 
of the Aluminum Research Laboratories. 



REFERENCES 

G. Tammann: A Textbook of Metallography, 33 

New York, 1925. The Chemical Catalogue C 
H. S. Strickler and H. Seltz: JnL Amer. Cher 

Soc. (1936) 58, 2084. 
J. A. McAteer and H. Seltz: JnL Amer. Cher 

Soc. (1936) 58, 2081. 
H. Seltz and B. J. Dewitt: JnL Amer. Chem. So 

(1938) 60, 1305. 
G. N. Lewis and M. Randall: JnL Amer. Chen 

Soc. (1921) 43, 433, 
Allmand and Ellingham s Applied Electrc 

chemistry: Chap. XIII Primary Cells. 
Lehfeldt s Electrochemistry: Chap. Ill Theor 

of Chemi-Electromotive Force. 
G. N. Lewis and M. Randall: Thermodynamie 

Chap. XXIX Galvanic Cell. 
L. C. Bannister and U. R. Evans: JnL Chem. So< 

(1930) 1361. 
U. R. Evans, L. C. Bannister and S. C. Brittor 

Proc. Roy. Soc, (1931) I3*-A, 355- 
R. M. Burns: The Bell System Tech. JnL (193^ 

15, 20. 
R. H. Brown and R. B. Mears: Trans. Electro 

chem. Soc. (1938) 74, 405. 
E. H. Dix, Jr. and H. H. Richardson: Tram 

A.I.M.E. (1926) 73, 560. 
A. J. Bradley and H. J. Goldschmidt: Monthl 

JnL Inst. Metals (1939) 6, 195. 



DISCUSSION 

(C. S. Barrett presiding) 
C. H. SAMANS,* State College, Pa. Oi 
page ri8 the authors say " ... it should b< 
possible to determine the amount of copper ir 
an aluminum-copper solid solution." It wouk 
be of interest to ascertain whether the methoc 
would be of any value in determining the cause 
of precipitation-hardening in these alloys 
whether it is due primarily to true precipitatior 
or to some pre-precipitation process. It would 
seem that if the former is true the copper atoms 
would no longer be in solid solution, hence the 

* Associate Professor of Metallurgy, The Pennsyl 
vania State College. 



DISCUSSION 



123 



method should be applicable regardless of 
the size of the CuAl 2 particles. However, in the 
latter case the copper atoms presumably would 
still be part of the solid solution even though 
they might be concentrated in regions that no 
longer have its exact symmetry. Would the 
potentials for the two cases differ? 

C. S. BARRETT,* Pittsburgh, Pa. In draw 
ing conclusions from irreversible potential 
curves of solid solutions during decomposition, 
it would seem to be necessary first to eliminate 
the possible effect of strains on the curves. Is it 
possible that strains generated in the matrix 
or in the precipitate during aging could account 
for any features in the aging curves of Fig. 1 1 ? 

R. H. BROWN (author s reply) .Answering 
Professor Samans: The data given in the paper 
show clearly that precipitation can be deter 
mined by potential measurements, provided the 
precipitation has gone far enough to sub 
stantially change the copper concentration 
of the anodic areas. However, his question 
undoubtedly refers to the early stages of age- 
hardening, and during these early stages there 
would be no significant change in the concen 
tration of copper in anodic areas, regardless 
of whether the copper particles that obstruct 
slip are true precipitate or some hypothetical 
agglomerate. In other words, the method 
described in the paper would not be expected 
to detect the early stages of precipitation. This 
has been borne out by experiment, since we 
have been unable to determine any change in 



* Metals Research Laboratory, Carnegie Institute 
of Technology. 



heat-treated and quenched duralumin by 
room-temperature aging. 

If a solution were found in which the anode 
areas would polarize readily and the cathode 
areas would not polarize at all, there might be 
some chance of detecting precipitation in the 
early stages. Even then, however, there is some 
doubt, because we do not know how much 
effect on potential would be expected, as 
copper atoms change from a random distribu 
tion (which they are supposed to have in knots 
and other hypothetical agglomerates) to the 
ordered arrangement they have in the precipi 
tate. That is, since the potential seems to be 
mainly dependent upon the concentration of 
the copper, it is doubtful whether changing 
from random distribution of copper atoms 
(knots) to an ordered arrangement of copper 
atoms (precipitate) would alter the potential 
measurably; assuming, of course, the same 
percentage of copper in each case. 

In reply to Professor Barrett s question con 
cerning the effect of strains on the measured 
potential, it would appear that if strains in the 
alloys are to affect the potential, a solution 
other than the one used in this investigation 
will be necessary. In the salt-peroxide solution, 
high-purity aluminum and aluminum-copper 
solid solutions, cold-worked to various degrees, 
were found to have the same potentials as 
annealed high-purity aluminum and rapidly 
quenched aluminum-copper solid solutions, 
respectively. 

It should be pointed out that the method is 
relatively new, and consequently all its limita 
tions, as well as many of its useful applications, 
have not been found. 



X-ray Study of the Solid Solubility of Lead, Bismuth and Gold 

in Magnesium 

BY FRANK FOOTE* AND E. R. JETTED MEMBERS A.I.M.E. 



(Cleveland Meeting, October 1940) 



PRECISION lattice-constant measurements 
have been widely used in the study of 
cubic solid solutions but as yet have been 
rarely applied to noncubic solid solutions. 
(See, however, references i and 2.) With 
the high precision at present attainable 
in the measurement of lattice constants, 
the X-ray method should be readily 
applicable to noncubic solid solutions. In 
the work reported here, hexagonal close- 
packed magnesium was used as the solvent 
and either lead, bismuth or gold as the 
solute. In the magnesium-lead and mag 
nesium-bismuth systems the a and c 
lattice constants were determined as a 
function of composition, densities were 
calculated from these lattice-constant data, 
these theoretical densities were compared 
with those directly measured and the solid 
solubility limit was determined on quenched 
specimens over a considerable range of 
temperature. In the magnesium-gold sys 
tem, the solid solubility is very small but 
we have obtained an indirect check on the 
solubility limit previously reported. 

MATERIALS AND PREPARATION or THE 
ALLOYS 

The magnesium used was obtained from 
the Aluminum Company of America and 
was 99.987 per cent pure. Impurities as 
reported on the label were: Fe, Al, 0.004 

Manuscript received at the office of the Institute 
July 25, iQ40. Issued as T.P. 1248 in METALS TECH- 
NOLDGV, December 1940. 

* Instructor in Metallurgy, Cooper Union, New 
York, N. Y. 

t Professor of Metallurgy, School of Mines, 
Columbia University, New York, N. Y. 

1 References are at the end of the paper. 



per cent; Si, 0.009 P er cent; Zn, trace; Cu, 
Pb, nil. The bismuth was Merck s analyt 
ical grade, minimum purity 99.95 per cent. 
Maximum impurities as listed on the label 
were: Fe, 0.008 per cent; Pb, o.oio per cent; 
Ag, o.oio per cent; Cu, 0.005 per cent; 
Sn, 0.003 P er cent > Sb, o.oo ; As, o.ooo; S, 
o.oi per cent. The gold was purified by 
double precipitation with sulphur dioxide. 
The lattice constants of these three metals 
have been previously reported. 3 The lead 
used was test lead, silver free and with a 
maximum of o.io per cent heavy metal 
impurities. The lattice constant of this 
lead has been measured and found to be 
4.94057 o.oooiB at 25C. 

The alloys were prepared by melting 
together weighed amounts of the constit 
uent metals in magnesia-lined alundum 
crucibles under an atmosphere of hydrogen 
purified by passing over magnesium heated 
to 500 C. The magnesia lining prevented 
contamination of the alloys with aluminum. 
Vacuum melting could not be used because 
of the volatility of magnesium. A high- 
frequency induction furnace was used for 
the melting operation and the stirring 
action of the current resulted in thorough 
mixing of the constituents. The melts were 
allowed to freeze in the crucible, the ingots 
obtained weighed about 10 grams. An 
extremely thin white film formed on the 
surface of the melts, sometimes covering 
only a part of the exposed surface. The 
ingots were filed clean, placed in small iron 
crucibles with close-fitting lids and indi 
vidually sealed into Pyrex glass tubes. 
Before sealing off, the tubes were alter- 



124 



FRANK FOOTE AND E. R. JETTE 



nately evacuated and filled with pure 
hydrogen a number of times to displace 
air. Finally hydrogen was admitted to a 
pressure of approximately half an atmos 
phere and the tube was sealed off. The 
ingots were annealed for at least a week at 
temperatures close to the eutectic tempera 
ture. This technique hinders the transfer 
of magnesium from the alloy to the glass 
container; the iron crucible prevents direct 
contact between the alloy and the glass, 
the hydrogen retards the diffusion of 
magnesium vapor. The alloys were then 
quenched in water and again cleaned by 
filing. The ingots were preserved in a 
desiccator. 

LATTICE-CONSTANT MEASUREMENTS 

Samples for X-ray analysis were prepared 
by filing. Two techniques were used in heat- 
treating these filings. For annealings below 
4ooC., the samples were sealed into 
evacuated glass tubes. At temperatures 
above 4ooC. these high-magnesium alloys 
rapidly attack glass, and films from such 
samples often showed diffuse lines. For 
many of these higher-temperature anneals 
the filings were packed into small thin- 
walled iron crucibles with loosely fitting 
lids. These crucibles were then sealed into 
glass tubes. The samples were annealed for 
various periods of time at suitable tempera 
tures in vertical rapid quenching furnaces. 
The samples were quenched by crushing 
the tubes under butyl acetate. A number of 
analyses showed no increase in iron content 
during this treatment. 

The general X-ray technique has been 
previously described. 3 The K a radiations 
from an iron target were used, the wave 
lengths were taken from the second edition 
of Siegbaum s book. 4 Five doublets were 
measured on each film. Correction for the 
index of refraction was applied and the 
lattice constants calculated by the method 
of Cohen. 6 3 All lattice constants were 
reduced to a common temperature of 
25C., using the expansion coefficients for 



magnesium as given by Raynor and 
Hume-Rothery. 6 

Many of these alloys are rapidly attacked 
by water or moist air, particularly with the 
sample in the form of filings. In dry air 
the alloys are stable, the ingots kept in a 
desiccator remained bright and clean indef 
initely. To minimize the effect of water and 
water vapor, the alloy filings were quenched 
in a nonaqueous liquid and the work was 
carried out during the cold winter months 
when the water-vapor content of the 
atmosphere was low. The mounted samples 
for X-ray exposure were prepared with 
Duco cement and a thin smear of the 
cement was placed over the sample. 
During the exposure, therefore, the filings 
were embedded in Duco cement and were 
not in direct contact with the atmosphere. 
The powders remained bright throughout 
the exposure. In all cases, sharp diffraction 
lines, indicating uniformity of composition, 
were obtained. 

DENSITY MEASUREMENTS 

Density measurements were made on 
many of the annealed single-phase alloys. 
The loss-of-weight method was employed, 
using monobrombenzene as the density 
liquid. This method has been previously 
described. 7 The density samples were an 
nealed and quenched filings, millings or 
small chips. No reaction between the 
alloys and the monobrombenzene was 
observed; the samples remained clean and 
bright throughout the determinations. 

CHEMICAL ANALYSIS 

The alloys were analyzed for lead, bis 
muth or gold following the methods given 
in Hillebrand and Lundell. 8 Lead was deter 
mined as Pb02 by electrolysis, bismuth 
as BiOCl and gold as free metal by precipi 
tation with sulphur dioxide. The methods 
were checked by analysis of solutions 
containing large amounts of magnesium 
and known amounts of the second metal 
Unless otherwise indicated, the analyses 



126 



SOLID SOLUBILITY OF LEAD, BISMUTH AND GOLD IN MAGNESIUM 



were made on a portion of the annealed fil 
ings used for the X-ray analysis. The den 
sity samples were analyzed after the density 
measurements were completed, the whole 



weight of the alloy and V f the volume of 
the unit cell in terms of cubic Crystal 
Angstroms (i Crystal Angstrom = i A cr = 
io 3 (X units). The quantity = io 24 //W 



C. 

5235 



a 

3205 
31OO 






n^-2cr 










- g 8 -" 










r 













O / ^ 3456? 

Atomic PerCenT Lead 

FIG. i. LATTICE CONSTANTS or SINGLE-PHASE MAGNESIUM-LEAD ALLOYS. 



sample being used for this analysis. All 
compositions were converted to atomic per 
cent, using for the atomic weights: Mg, 
24.32; Pb, 207.21; Bi, 209.00; Au, 197.2. 

CALCULATION OF THEORETICAL DENSITY 

Jette and Foote 9 have shown that the 
theoretical density should be calculated 
from lattice-constant data by means of 
the following formula: 

_ IQ24 ^M v- n ^ 
P = /w T - A ~7/~ [i] 



has been evaluated from published data 
on calcite and was found to have a 
numerical value of 1.6 50 23.* Magnesium 
has a hexagonal close-packed structure, 
contains two atoms per unit cell and the 
volume of the unit cell can be calculated 
from the a and c lattice constants by the 
equation V J*j\/3 a2 - Introducing these 
constants into equation i, our density 
formula becomes: 



1.65023-2-2 M 
~ ~ 



where N is Avogadro s number, / the ratio 
between the absolute and the relative 
scale of X-ray wave lengths, n the number 
of atoms per unit cell, M the average atomic 



p = 



M 
= 3.81104 X -jr. grams/cu. cm. [i ] 

U/ 

It should be emphasized that in equation 
i the value of M depends upon the type 



FRANK FOOTE AND E. R. JETTE 



127 



of solid solution assumed and that a and 
c are expressed in terms of " Crystal 
Angstroms." 

LATTICE CONSTANTS AND DENSITIES OF THE 

SINGLE-PHASE MAGNESIUM-LEAD AND 

MAGNESIUM-BISMUTH ALLOYS 

The lattice-constant data for the single- 
phase magnesium-lead alloys are tabulated 
in Table i and plotted in Fig. i. Similar 



the compositions were obtained by analysis 
of a portion of the annealed sample used 
for the X-ray analysis. The average stand 
ard error of the a lattice constants was 
0.00020, for the c lattice constants 0.00017 
and for the ratio c/a, 0.00012. Both lattice 
constants increase with increasing alloy 
content, the a constant slowly, the c 
constant much more rapidly. Hence, the 
c/a ratio increases with increasing solute 



TABLE i. Singk-phase Magnesium-lead Alloys 



Alloy 


Heat Treament 


At. Per Cent 
Pb . 

(Analysis) 


Lattice Constants (25C.) 


Hours 


Deg. C. 


a 


c 


c/a 


]Vlg 


(Average of 6) 
3 
119 
5 
3 
119 
5 
117 
5 
3 
117 
4-5 
118 
6 
4 5 
4-5 


(Ref. 3) 
403 
410 
448 
403 
410 
439 
400 
448 
438 
400 
450 
400 
449 
448 
448 


00 

0.51 
0.53 

I. 00 
I . 21 
I. 21 
I 93 
2 00 
2 32 
2.73 
2.84 
3 59 
3-73 
4-37 
5-07 

6.20 


3-2030 
3.20 3 3 
3 2033o 
3.2036 
3 2044 
3 2040 
3 2046" 
3-2047^ 
3.2052^ 
3 2060^ 
3 2 055o 

?:& 

3.20684 
3-20690 
3-2074 4 


5 2002 
S-20I9" 
5-2023^ 
5-2043J 
5.20536* 
5.2052 
5.2082;? 
5.208s: 
5-2100^ 

5.2125, 

5.2127 
5.2168? 
5.2172 

5 2200;; 
5 2247* 

5 2316^ 


1.6235 
.62394 
.6240^ 
.6245* 
.62444 
62464 
6252" 
62 52 4 
.6254* 
6258$ 
6261* 
6274? 
.6270* 
62 7 9g 
.6291 
.6310^ 


15 

I ga 


18 ... 


IA 


IAS 


II. ... .... 


I IS 


ip 


10 ... ... 


I OS 





I2g 


2O ... ... 


16 


17 . 





a Samples marked g were annealed in glass tubes, all others were annealed in iron crucibles. 

TABLE 2. Single-phase Magnesium-bismuth Alloys 



Alloy 


Heat Treatment 


At. Per Cent 
Bi 
(Analysis) 


Lattice Constants (2sC.) 


Hours 


Deg. C. 


a 


C 


c/a 


Mg 


(Average of 6) 

I.O 

0.5 
0.5 

I.O 

o.S 
o 5 

I.O 
I.O 

0.5 
0.5 

I.O 
I.O 
2 
2.0 
I 


(Ref. 3) 
556 
536 
536 
547 
536 
536 
541 
547 
532 
532 
547 
541 
550 
553 
548 


0.00 

o. 24 
(0.25) 

(0.25) 
0.46 
(0.50) 
(0.50) 
0-57 
0.58 
(0.76) 
(0-76) 
0.80 

0.81 

(i 00) 
I 00 
I 01 


3-2030 
3.2032^ 
3-2032" 
3-2034J 
3.2036:, 
3.2042;: 
3.2042 
3. 2035 J 
3 2034; 
3.2049. 
3-2045^ 
3-2041 
3-2042g 
3.2043:: 
3.2042^ 

3.20 S 2 6 


5-2002 
5-2018" 
5.2017 

5.2018^ 
5-2035o 

5.2042 

5.2042? 
5 - 2045" 
5.2047, 
5 . 2060^ 
5-2058? 
5-2059, 
5.2058^ 
5-2072^ 
5 2073? 
5-20745 


i.62 3 5 4 
1.62393 
1.6239 
1.6238? 

I.6242g 

1. 624i!J 
1.6241; 
1.6246? 
1.624?! 
1.6243^ 
1.6245 
1.6247! 
1.6246? 
1.6250 

1.6251* 
1.6256* 


20 


20 


2O 


24 . 


21 


21 . 


25 . 


25 , 


22 . . .... 


22 


22 ... 


22 . . . 


15 , . 


15 . 


15 .... 





Analyses in parentheses are ingot analyses. 



data for the magnesium-bismuth alloys are 
tabulated in Table 2 and plotted in Fig. 2. 
The heat -treatments refer to the final 
powder anneals. Unless otherwise stated, 



content and approaches but does not reach 
the theoretical value of 1.633 for hexagonal 
close-packed spheres. Raynor 1 found this 
same effect in the Mg-Ag, Mg-Ga and 



128 



SOLID SOLUBILITY OP LEAD, BISMUTH AND GOLD IN MAGNESIUM 



Mg-In systems. There are large negative 
deviations from Vegard s rule for the 
additivity of atomic volumes. The lattice 
constants as a function of composition can 



TABLE 3. Comparison of Calculated and 
Measured Densities 



c. 

06 



\S20Z 




Bismuth 

FIG. 2. LATTICE CONSTANTS OF SINGLE-PHASE 
MAGNESIUM-BISMUTH ALLOYS. 

be satisfactorily represented by the least- 
squares equations: 

Mg-Pb a = 3.20292 + 0.001104 (at. % Pb) 
- 0.000060 (at. % Pb) 2 [2] 

with an average deviation of 0.00018 A cr or 
0.23 at. per cent Pb. 

c = 5.20020 + 0.003902 (at. % Pb) 

+ 0.000188 (at. % Pb) 2 [3] 



Alloy 


At. Per 
Cent Pb 


P (25C.) 
(Calcu 
lated) 


P (2SC.) 
(Meas 
ured) 


Ap 


Mg . . 


o.oo 


I T\1 


I 736 




15 


o 60 


I 815 


i 807 


on 8 


18 


1 . 04 


I 870 


I 866 


. uuo 


14 


1 .24 


1.896 


i 888 


o 008 


19 .... 


2.36 


2 038 


2 O^O 




20 


4.25 


2. 279 


2 277 





16 . 


5 07 


2 382 


2 376 


nn A 


17 


6 26 


2r ?2 




. OO 














At. Per 
Cent Bi 








Mg. 


00 


I 7373 


I 7^62 




20 


o 25 


i 7689 


I 7666 




24. . 


o 46 


I 7954 




W . UUJS J 


25 


o 58 


I 8107 


8063 




25 


o 60 


I 8141 


8l4.I 




22. . . 


o 78 


I 8367 


877O 




26 


o 95 


i 858 1 


QcR/i 


} oo^s 













with an average deviation of o.oooio A cr 
or 0.022 at. per cent Pb. 

Mg-Bi a = 3.20298 + 0.001702 (at. % Bi) 

W 

with an average deviation of 0.00023 A cr 
or 0.14 at. per cent Bi. 

c = 5.20015 + 0.008310 (at, % Bi) 

- 0.001131 (at. % Bi) 2 [5] 

with an average deviation of 0.00016 A cr 
or 0.019 at - P er cent Bi. 

Theoretical densities can be calculated 
from these lattice-constant data by equa- 



TABLE 4. Two-phase Magnesium-lead Alloys 



Alloy 


Alloy Com 
positions, 
At. Per Cent 


Heat Treatment 


Lattice Constants (25C.) 


Solubility Limit, 
At. Per Cent Pb 












Pb 


Hours 


Deg. C. 


a 


c 


(Eq. 3) 


8 


7-3 


5-5 


463 


3.2099, 


5-2345 


6.67 


8 


7-3 
7-3 


4-5 
5 


450 
438 


3.2077. 
3-2076^ 


5.2326? 
5.2316^ 


6.36 

6.20 




7-3 


5 


438 


3 2077? 


5.2313^ 


6.15 






6 


402 


3-2071 


5.2253 


5- 17 


8 
16 


7-3 
5-J 


5 
122 


392 

345 


3-2o6 4 4 
3-2055^ 


5.2233 
5.2167-: 


4.83 
3.62 


log 


3-6 

2-7 


501 
501 


300 
300 


3-2051: 
3-2054! 


5.2113;: 

5.2113; 


2-55 
2.55 


142 


1.9 

1.2 


506 
736 


257 
198 


3-20415 
3-2041^ 


5.2073g 
5.2039, 


1.70 
0.92 






55i6 


152 


3.2033^ 


5.202i 


0.50 


fl Samples marked g were annealed in glass tubes, all others were annealed in iron crucibles. 



PRANK POOTE AND E. R. JETTE 



I2Q 



tion i . The calculation has been made on 
the assumption of simple substitution of 
lead or bismuth atoms for magnesium 



Table 3 gives the comparison between 
directly measured and calculated densities. 
The uncertainty in the calculated densities 




a 



36763/0 

PerCenf Lead 

FIG. 3. PHASE DIAGRAM OF MAGNESIITM-RICH MAGNESIUM-LEAD ALLOYS. 
Dashed line shows solid solubility limit obtained by Vossktihler. 12 



atoms in the magnesium lattice. On this 
basis, M is given by: 

For Mg-Pb ; M = 24.32 + 1.829 

(at. % Pb) [6] 

For Mg-Bi, M = 24.32 + 1.8468 

(at. % Bi) [7] 



is slightly less than o.ooi density unit, the 
greater part being derived from the uncer 
tainty in the atomic weight of magnesium 
(and the accidental error of the analysis). 
The uncertainty in the directly measured 
densities is slightly greater than o.ooi 
density unit, assuming that the uncertainty 



TABLE 5. Two-phase Magnesium-bismuth Alloys 



Alloy 


Alloy Com 
positions, 
At. Per Cent 
Bi 


Heat Treatment 


Lattice Constants (25C.) 


Solubility Limit, 
At. Per Cent Bi 
(Eq. 5) 


Hours 


Deg. C. 


a 


C 


16 
23 


2.00 

1.28 


I.O 
I.O 


548 
548 


3.2049, 
3 - 2049^ 


5.2077, 
5.2076:* 


1.07 
1. 05 


23 


1.28 


0.5 


544 


3-2048* 


5.2076? 


i. 05 


23 


1.28 


o.S 


544 


3 . 20449 


5-2079; 


1. 10 


IS 


I. 00 


I.O 


542 


3-2045. 


5-207I 2 


0.97 


23 


1.28 


I.O 


540 


3-2042;: 


5.2074^ 


1.02 


23 


1.28 


0.5 


538 


3.2045* 


5.2073^ 


I.OO 


18 


3 93 


2.5 


538 


3.2041* 


5.2064 


0.86 


IS 


I. 00 


i. 5 


524 


3-2045 


5-2059g 


0-79 


22 


0.75 


2.0 


493 


3.20432 


5.2046 


0.58 


16 


2 00 


I.O 


491 


3 2043 


5-2045 


0.57 


17 


3.00 


I.O 


489 


3- 2035 


5 2043^ 


0.55 


16 


2.00 


2.0 


453 


3-20333 


5.2029J 


0.35 


17 


3.00 


3-0 


453 


3.2032^ 


5-2028^ 


0-34 


22 


0.75 


8.0 


448 


3-2032;* 


5-2027^ 


0.33 


24 

16 


0.45 
2. 00 


35 

119 


425 
362 


3.2030 
3.2026 


5.2021^ 

5.2012:? 


0.25 
0.13 


IS 


I. 00 


119 


362 


3-2029,5 


5-2010^ 


0.12 


152 


I. 00 


5497 


307 


3-2028 


5 2004^ 


0.03 


16* 


2.00 


5497 


307 


3.2030 Z 


5.20055 


0.05 



Samples marked g were annealed in glass tubes, all others were annealed in iron crucibles. 



130 



SOLID SOLUBILITY OF LEAD, BISMUTH AND GOLD IN MAGNESIUM 



in each weighing is o.oooi gram. The 
combined uncertainty is approximately 
0.002 density units. In some cases the 
difference between directly measured and 




4 6 & SO / 

Atom/ c Per Cent Bismuth 

FIG. 4. PHASE DIAGRAM OF MAGNESIUM-RICH 

MAGNESIUM-BISMUTH ALLOYS. 
Solid circles indicate four points obtained by 
Grube, Mohr and Bornbak 17 on curve of solid 
solubility limit. 

calculated densities exceeds this combined 
uncertainty. In all cases of this kind the 
measured density is lower than the cal 
culated value and the lowered density can 
be ascribed to physical inhomogeneity; 
i.e., to the presence of cracks, blowholes, 
porosity, or other fault. However, the 
agreement is considered to be sufficiently 
good to establish the fact that both of these 
high-magnesium solid solutions are of the 
simple substitutional type. 

Tables 4 and 5 give the lattice-constant 
data on two -phase alloys and the solid 
solubility limits calculated from these data. 
The solubility limits were obtained by 
substituting the measured lattice-constant 
values in equations 2, 3, 4 or 5 and solving 



for atomic per cent Pb or at. per cent Bi. 
Since the a lattice constant changes only 
slightly with composition, the solubility 
limits obtained from equations 2 and 4 are 
not very accurate. The limits obtained from 
the c lattice constants (equations 3 and 5) 
are considerably more reliable. We have 
tabulated only the solubility limit values 
as obtained from the c lattice constants. 
In general the limits obtained from the a 
constants confirm the reported solubility 
limits, but are scattered in a band on both 
sides of the lines shown in Figs. 3 and 4. 

The solubility limit of lead in magnesium 
has been plotted in Fig. 3. Also plotted in 
Fig. 3 are the liquidus as drawn through 
the thermal points obtained by Grube, 10 
Kurnakow and Stepanow 11 and Voss- 
kuhler; 12 the solidus as obtained by 
Vosskiihler 12 from temperature-resistance 
measurements; and the solid solubility 
limit (dashed line) as obtained by Voss 
kiihler 12 from conductivity and from 
temperature-resistance measurements. The 
eutectic temperature was taken as 468 C., 
the average of the four previously reported 
values: 46oC., 10 466C., 12 47oC. 13 and 
475C. n 

Hansen 14 15 has reported a few micro 
scopic observations on the solid solubility 
limit. His alloys were given short-time 
anneals (48 hr.) and his results were not 
conclusive. 

Our solid solubility limit agrees reason 
ably well with that of VosskiihlerV 2 up to 
about 400 C. In this temperature the 
maximum difference between the two 
sets of results is approximately 0.3 at. 
per cent Pb, or i5C. Above 4ooC., the 
differences are much greater, reaching 
about 2.1 at. per cent Pb at the eutectic 
temperature. It is in this temperature 
region that the results of conductivity and 
temperature-resistance measurements are 
least reliable (see Vosskuhler s Figs. 2 and 
3)* In this same region, our c lattice con 
stant is changing most rapidly with com 
position; that is, it is in this region that 



DISCUSSION 



errors in measuring lattice constants have 
the least effect on the solubility limit. 
Further, our alloy No. 8, containing 7.3 at. 
per cent Pb and quenched from 450 and 
463C., is definitely two-phase. We believe, 
therefore, that our X-ray results are more 
reliable than those obtained by conduc 
tivity methods. 

The solubility limit of bismuth in mag 
nesium has been plotted in Fig. 4. Also 
plotted in Fig. 4 are the liquidus as drawn 
through the thermal points of Grube 16 and 
Grube, Mohr and Bornhak; 17 the solidus 
as determined by Grube, Mohr and Born 
hak 17 from temperature-resistance measure 
ments. The eutectic temperature was taken 
as 553C. 16 - 17 

Grube, Mohr and Bornhak 17 obtained 
four points on the curve of solid solubility 
limit. These points are shown on Fig. 4 as 
solid circles. All of these points are within 
0.15 at. per cent Bi of the curve drawn 
through our results. 

TABLE 6. Summary of Solubility Limit, 
Lead and Bismuth in Magnesium 



t, 
Deg. C. 


At. Per 
Cent Pb 


Wt. Per 
Cent Pb 


At. Per 

Cent Bi 


Wt. Per 
Cent Bi 


553 






1.15 


9.05 


525 






0.8i 


6.56 


500 






0.60 


4-95 


468 


7.16 


39-6 






450 


6.54 


37-5 


0.34 


2.81 


400 


4-99 


30-9 


0.18 


1-53 


350 


3.64 


24.3 


0.09 


0-79 


300 


2.51 


18.0 


0.04 


0.33 


250 


l.6l 


12.3 


O.OI & 


0. IO & 


200 


0.95 


7.5 






150 


0.49 


4.0 






100 


0.2I 6 


I.8& 







Eutectic temperature. 
6 Extrapolated value. 

Table 6 gives a summary of the solid 
solubility limits of lead and bismuth in 
magnesium. These values were read from 
large-scale plots. 

Magnesium-gold Alloys 

Hume-Rothery and Butchers 18 have 
shown by microscopic methods that the 
solubility of gold in solid magnesium is 
approximately o.i at. per cent at the 



eutectic temperature. This slight solubility 
makes the establishment of a lattice-con 
stant composition curve difficult. An alloy 
containing 3 at. per cent Au was annealed 
4 hr. at 56oC. and quenched. The alloy 
was two-phase, the lattice constants of the 
magnesium-rich phase were found to be: 
a = 3.20216, c = 5.19924 at 25C. Gold 
in solid solution in magnesium decreases 
both lattice constants, a by 0.00084 and 
c by 0.0009? A cr . These changes in lattice 
constants are consistent with a solid solu 
bility of approximately o.i at. per cent Au 
as found by Hume-Rothery and Butchers. 18 
This conclusion was reached by an indirect 
method; further work is planned. 



REFERENCES 

1. G. V. Raynor: Proc. Roy. Soc. London (1940) 

1 74 -A, 457-471. 

2. W. Hume-Rothery and G. V. Raynor: Proc. Roy. 

Soc. London (1940) I74-A, 471-486. 

3. E. R. Jette and F. Foote: Jnl. Chem. Phys. (1935) 

3, 606-616. 

4. M. Siegbaum: Spektroskopie der Rontgenstrahlen, 

Ed. 2, Berlin, 1931. Springer. 

5. M. U. Cohen: Rev. Sci. Instruments (1935) 6, 

68-74. 

6. G. V. Raynor and W. Hume-Rothery: Monthly 

Jnl. Inst. Metals (1939) 6, 477-485. 

7. F. Foote and E. R. Jette: This volume, page 152. 

8. W. F. Hillebrand and G. E. F. Lundell: Applied 

Inorganic Analysis. New York, 1929. John 
Wiley and Sons. 

9. E. R. Jette and F. Foote: Phys. Rev. (July 1940). 

10. G. Grube: Ztsch. anorg. Chem. (1905) 44, 117-130. 

11. N. S. Kurnakow and N. J. Stepanow: Ztsch. 

anorg. Chem. (1905) 46, 177-192. 

12. H. Vosskuhler: Ztsch. Metallkunde (1939) 31, 

109-111. 

13. E. Abel, O. Redlich and F. Spausta: Ztsch. anorg. 

allge. Chem. (1930) 190, 79-89. 

14. M. Hansen: Ztsch. Metallkunde (1927) 19, 

455-456. 

15. M. Hansen: Aufbau der ZweistofHegierungen, 

865. Berlin, 1935. J. Springer. 

16. G. Grube: Ztsch. anorg. Chem. (1906) 49, 72-92. 

17. G. Grube, L. Mohr and R. Bornhak: Ztsch. 

Elektrochem. (1934) 40, 143-150. 

1 8. W. Hume-Rothery and E. Butchers: Jnl. Inst. 

Metals (i937) 60, 345-350. 

DISCUSSION 

(P. H. Brace presiding) 

G. EDMUNDS* AND M. L. FULLER,* Palmer- 
ton, Pa. We disagree with the opening state 
ment made by the authors that precision 
lattice measurements have been rarely applied 
to noncubic solid solutions. This statement is 
readily refuted by the literature on the subject; 
e.g., see reference 15, which refers to numerous 

* Research Division, The New Jersey Zinc Co. 
(of Pa.). 



132 



SOLID SOLUBILITY OE LEAD, BISMUTH AND GOLD IN MAGNESIUM 



publications giving lattice-parameter deter 
minations on noncubic metals significant to 
four and more figures. 

For the determination of solid solubilities in 
some alloy systems, very much more precise 
measurements are needed than in others to 
obtain a given accuracy. In any case proper 
specimen preparation and analyses are essen 
tial, as has been recognized in the present 
work. 

Lattice constants are reported in the paper 
to six figures, the sixth figure being given in 
subscript position. This creates the impression 
that the lattice constants are correct to five 
figures. We suggest that five figures would have 
been adequate, since the standard errors 
calculated by the authors from each X-ray 
photogram average two units in the fifth 
figure. That the accuracy claimed is too high 
becomes more obvious from an inspection 
of the curves of Fig. 2. The curves of that 
figure show the variation of the lattice con 
stants a and c with composition. The a values 
deviate widely from the curve this deviation 
amounting to as much as six units in the fifth 
figure of the value of a. Less deviation is noted 
for the c values. The authors do not explain 
this difference in the accuracy of a and c. 
If the a values are inaccurate, and since both a 
and c are calculated from the same original 
diffraction-line measurements, how reliable 
are the c values? 

It seems that the accuracies of the lattice 
constants of the magnesium-bismuth alloys 
and the values of calculated densities and solid 
solubilities derived therefrom are open to 
serious doubt, because of the striking lack of 
conformity to a smooth curve of the values of 
Fig. 2. Because the a values are squared, errors 
in them have a larger effect on calculated 
densities than do errors in c values. 

F. FOOTE AND E. R. JETTE (authors reply). 
We believe that our introductory statement 
is essentially correct. We do not consider that 
any lattice-constant measurement may claim 
accuracy of more than four significant figures 
unless the systematic errors have been removed. 
These ordinarily are considerably larger than 
the accidental errors. While methods of elimi 
nating systematic errors for cubic systems 
have been well known for many years, no really 
satisfactory method for noncubic systems was 



published prior to 1935 (ref. 5), the year 
Hanson (ref. 15) was printed. Unless the 
"drift" (cf. refs. 5 and 3) is accounted for, 
the fifth significant figure is highly uncertain 
and sometimes even the fourth. 

The vexing question of just how many 
figures should be carried in stating a result still 
lacks a completely satisfactory answer. It 
should be noted that the errors quoted are 
"standard errors," which may be from 50 to 
several hundred per cent greater than the 
ordinary "probable errors." In some of our 
earlier articles an even more conservative 
estimate of the accuracy of our measurements 
was used; namely, a fiduciary limit (ref. 3). It 
is to be emphasized that the standard error in 
a and c was calculated from the actual measure 
ments and not merely estimated by some 
scheme such as taking the error in the line 
position as 0.05 mm. Such standard errors 
carry the correction factor for the number of 
observations made; i.e., the number of line 
pairs on each film. Such care and conservatism 
in stating the errors of lattice-constant measure 
ments is ordinarily unnecessary and, accord 
ingly, most authors do not use such methods. 
In any critical comparison of measurements 
from different sources, the method used in 
calculating the errors is nearly as important as 
the method of making the measurement itself. 
Since the conversion from one method of stating 
errors to another involves more significant fig 
ures than would be required in the simple rules, 
based in turn on probable errors, we have 
adopted the form used in these articles. It is 
clear and should lead to no misunderstanding. 

The wave lengths used for the lattice-con 
stant measurements were chosen so as to give 
lines suitable for the accurate evaluation of the 
c constants. This procedure is entirely justified 
by the facts that: (i) the constant that changes 
the more rapidly with composition is obviously 
the proper one to use for determining solubility 
limits (Figs, i and 2 show that c changes much 
more rapidly than a), and (2) the error in the 
density contributed by the a constants is small 
compared with that due to the uncertainties in 
atomic weights (see below). The conditions that 
must be satisfied to obtain equal percentage 
errors in a and c were given in ref. 3. For the 
reasons given above, they were deliberately 
violated so as to favor a higher accuracy for c. 



DISCUSSION 



133 



The error in the calculated densities is cor 
rectly stated; the methods used in the calcula 
tion are completely covered in ref . 9. The error 
due to uncertainties in the lattice constant are 
negligible in comparison with those due to 
uncertainty in the atomic weights. The average 
lattice-constant errors could be at least twice 
as large as they are and in no case would the 
errors in the calculated density be as great as 
that contributed by the uncertainty in the 
atomic weights, the latter uncertainty being 
taken as plus or minus one unit in the last 



quoted figure for the atomic weight as given 
in the official table for 1940 (cf. ref. 9), 

The points on the solubility curve are 
affected not only by the errors in measuring 
the lattice constants but also by the degree of 
completeness with which equilibrium was 
reached during annealing and by retention of 
this state during quenching. This discussion 
of the errors and the " fit" of the points in the 
solubility curves (Figs. 3 and 4) should answer 
any questions as to the accuracy of these 
curves. 



Mechanism of Precipitation from the Solid Solution 
of Silver in Aluminum 

BY C. S. BARRETT,* MEMBER, A. H. GEISLER,! STUDENT ASSOCIATE, AND R. F. MEHL,* 

MEMBER A.I.M.E. 

(New York Meeting, February 1941) 



THE complicated nature of the property 
changes that accompany age-hardening 
has made it necessary to reconsider and to 
elaborate the simple dispersion theory. 1 It 
has been apparent for some time that more 
direct information is needed concerning 
the atomic mechanism of precipitation 
from solid solution if the theory of age- 
hardening is to be developed past a purely 
qualitative stage. J Fortunately, in recent 
years information has been developed 
which allows us to trace in detail the 
lattice changes accompanying tne process 
of precipitation from solid solution. This 
information relates chiefly to the sequence 
of the several stages of lattice alteration 
by which the matrix lattice is transformed 
to the equilibrium precipitate lattice, and 
permits reasonable inferences to be drawn 
concerning the states of internal strain that 
accompany these stages. 

This information is most complete for 
the solid solution of copper in aluminum. 
The equilibrium precipitate is the 6 phase, 
"CuAl*" This forms in plates, parallel 
to the jiooj matrix planes. The atom 
pattern and atom spacing on the plane 
in 6 lying parallel to this plane are roughly 
similar to those opposite it in the matrix. 3 
The formation of 0, however, is preceded 

Manuscript received at the office of the Institute 
Nov. 15, 1940, Issued as T.P. 1275 in METALS 
TECHNOLOGY, February 1941. 

* Metals Research Laboratory, Department of 
Metallurgy, Carnegie Institute of Technology, Pitts 
burgh, Pa, 

f Aluminum Company of America Graduate Pel- 
low, Department of Metallurgy, Carnegie Institute 
of Technology, Pittsburgh, Pa. 

1 References are at the end of the paper. 

J Studies of the crystallography of Widman- 
statten figures have had this as their chief aim. 2 



by the formation of a transition lattice , 
which again lies parallel to the matrix {100} 
plane, but in this case the matching of 
atom pattern and spacing across the 
matrix-precipitate interface is nearly per 
fect; thus there is lattice coherency across 
the interface and the interface is a plane 
of separation in only a very limited sense. ^ 8 
Particles of may be recognized micro 
scopically and form a well-defined Wid- 
manstatten figure.* 

The lattice is recognized by the occur 
rence of diffraction lines on powder photo- 
grams or of spots on Laue photograms 
which may be analyzed in the usual ways 
to determine the lattice type and the 
lattice spacings. Very recent research 6 - 7 - 1 
has shown, however, that streaks occur on 
the Laue photogram at an earlier point in 
time during the aging process. These 
streaks have been attributed to a prefer 
ential clustering of copper atoms into 
small platelike elements of only a few 
atoms in thickness and a few hundred 
atoms in diameter, probably of the compo 
sition of " CuAl 2 ," which may he designated 
as Guinier-Preston zones or aggregates. 1 
It is difficult if not impossible to determine 
the lattice type and lattice spacing in these 
aggregates. It is possible 1 that these 
aggregates are in fact merely 6 in a high 
state of dispersion; the inclination to argue 
this is strong, as we shall see later. Never- 



* It was foreseen by Merica that restraints 
to lattice movements imposed by lattice regis 
try would lead to the formation of transition 
lattices. 9 



134 



C, S. BARRETT, A. H. GEISLER AND R. F. MEHL 



135 



theless, in the interests of conservatism, 
these aggregates will be designated as a sep 
arate stage in the process of precipitation. 

These aggregates in the aluminum-cop 
per system are again parallel to the (100) 
matrix planes. It appears that the harden 
ing is associated with either the Guinier- 
Preston aggregate or or both, 4 ~ 8 and 
that this hardening originates in the high 
degree of strain between these and the 
matrix. The equilibrium lattice 6, how 
ever, contributes relatively little to hard 
ness, for in this case accurate registry 
across the interface does not obtain; the 
precipitate particle has broken away from 
the matrix with the release of strain. 
When this occurs in the aluminum- copper 
system, the 6 particle (formed only at 
high temperatures on long-time aging) is 
already too large to be considered as 
contributory to hardening on the basis 
of the original age-hardening theory. It is 
to be noted that recent studies have thrown 
some doubt upon the conception of a 
critical dispersion of precipitate for maxi 
mum hardening, 10 - 12 and that accordingly 
this point may not be significant. Approxi 
mate values of the stress necessary to 
maintain lattice registry across the inter 
face have been calculated. 

The sequence of stages in precipitation 
in this system is accordingly as follows: 
matrix, Guinier-Preston aggregates, , and 
finally B. The precipitation reaction is 
thus in fact a series of consecutive reac 
tions. These stages are related, however, 
not only in a genetic sense that the 
transition lattice is formed from the 
Guinier-Preston aggregate, and the equi 
librium lattice from the transition lattice, 
but also in a geometrical sense, for each is 
in the form of plates parallel to the same 
matrix plane and has an orientation closely 
allied to that of the preceding stages. This 
similarity in geometry suggests strongly 
that the three stages represent only differ 
ent stages in a process of alteration in 
lattice geometry, the several stages differ 



ing only in the degree of completion of this 
alteration. 1 * 

This argument is based on one example 
only. It might well be argued that the 
occurrence of a common conjugate or 
interface plane throughout the three 
consecutive stages may be fortuitous and 
should not be used as the basis of a general 
theory. If, however, a system could be 
chosen that forms a platelike Widman- 
statten figure on some matrix plane other 
than the {iooj, and if it could be shown 
that the Guinier-Preston aggregate (and 
perhaps a transition lattice also) forms on 
this same plane, it could be more urgently 
argued that the Guinier-Preston aggregate 
is but an early stage in the development 
of the final equilibrium lattice, in which 
the lattice movements that ultimately will 
convert the matrix lattice to the equi 
librium lattice have already begun. 

An opportunity to test this argument is 
offered by the precipitation of the 7 phase 
from the solid solution of silver in alumi 
num. Studies of the Widmanstatten figure 
in this system 2 have shown that the close- 
packed hexagonal intermediate solid solu 
tion 7 forms plates parallel to the {111} 
plane in the matrix. On the basis of the 
argument, we may predict, therefore, that 
the Guinier-Preston aggregate and a 
transition lattice, if formed at all, will be 
plate-shaped elements lying parallel to 
this same plane. 

The present paper will present evidence 
for the correctness of this prediction. The 
alloy forms the Guinier-Preston aggregate, 
consisting of plates of which the orienta 
tion and approximate sizes have been 
determined; a transition lattice occurs, the 
structure and the orientation of which 
have been solved; from these data the 



* In this view it is as justified to designate 
the formation of the Guinier-Preston aggregate 
as precipitation as it is so to designate the 
formation of either of the latter two stages. 1 
Although this is a debatable point, we shall use 
the word "precipitation" in this sense in the 
present paper. 



136 PRECIPITATION FROM THE SOLID SOLUTION OP SILVER IN ALUMINUM 



lattice shifts necessary to transform the 
matrix lattice through the several stages 
to the equilibrium lattice have been 
derived; the nature and the approximate 
magnitude of the stresses associated with 
the formation of the transition lattice, 
and the release of these stresses on the 
formation of the equilibrium lattice, have 
been studied. 

THE SYSTEM ALUMINUM-SILVER 

Silver dissolves in solid solution in 
aluminum up to 48 weight per cent at the 
eutectic temperature SS8C. This solid 
solubility decreases to nearly zero at 2ooC. 
The alloy with 20 weight per cent Ag 
used in this study crosses the solvus curve 
on cooling at just under 45oC. The solid 
solution of silver in aluminum, hereinafter 
designated as d, is in equilibrium below the 
eutectic temperature with the 7 phase. The 
7 phase is an intermediate solid solution 
containing 85 to 90 weight per cent Ag. 
The composition of 7 in equilibrium with d 
does not vary with temperature. 13 

The d phase is face-centered cubic and 
is a substitutional solid solution. Owing 
to the near identity of the atomic radii of 
aluminum and silver, the side of the unit 
cube does not vary with concentration, 
retaining the value for aluminum, 4.041 A. 
The 7 phase is close-packed hexagonal, 
with the atoms distributed at random, and 
with a = 2.879, o = 4-573, and c /#o 
1.588, when saturated with aluminum 
(85 weight per cent Ag). 13 

Well-defined Widmanstatten figures are 
formed on the decomposition of the d 
phase. 2 It has been shown that these take 
the form of plates parallel to the { 1 1 1 } 
d planes. The orientation relationship is 
\ni\d || {00.1)7, and [ilo] || [21.0]. Thus 
the basal planes of the hexagon in 7 lies 
parallel to the octahedral plane in 5; the 
pattern of atoms in each is hexagonal. 
These hexagons are simply superimposed 
at the interface with the close-packed 
directions parallel. 2 



Until recently it has been assumed that 
the precipitate on aging is the 7 phase. 
Recently the occurrence of a transition 
lattice has been shown, though its struc 
ture has not been determined. 14 This 
lattice we shall designate the 7 lattice. It 
appears that the hardening on aging is 
associated with the formation of this lattice 
and not the 7 lattice. 14 

Preparation of Samples 

An alloy of 20.2 weight per cent Ag 
was prepared by melting 99.97 per cent 
Al with 99.99+ per cent Ag in a graphite 
crucible under a potassium chloride flux, 
chill-casting and rolling to a reduction of 
about 85 per cent with intermediate 
anneals. The alloy was annealed 48 hr. at 
55oC. for homogenization. From this 
stock, wires were drawn with reduction 
from 0.25 to 0.03 5-in diameter with two 
intermediate anneals to serve as samples 
for microscopic studies and for X-ray 
powder-diffraction studies. From the same 
stock large grains for Laue samples were 
grown in J^-in. square rods, by a strain- 
anneal method; individual crystals were 
cut out with a jeweler s saw, and carefully 
thinned by rubbing on emery paper and 
by deep etching, which in addition to 
thinning served also to remove strained 
metal. The aging treatments employed are 
listed below. 

Powder Photograms Experimental 

The wire specimens described above 
were given a solution heat-treatment of 
approximately i hr. at 500 to 5$oC. in 
evacuated glass tubes, and then quenched 
in ice water, upon which the tubes broke. 
One set of powder specimens (those used 
for aging at i73C.) was prepared by 
filing; the powder was soaked 20 min. in an 
evacuated tube at so6C., and quenched 
in ice water. Both wire and powder 
specimens, protected by enclosure in 
evacuated glass tubes, were placed in an 
aluminum block in a resistance furnace 



C. S. BARRETT, A. H. GEISLER AND R. F. MEHL 



137 



controlled to iC., or immersed in a 
salt bath controlled to KC. for aging 
treatments. 
Powder photograms were made in a 



be seen by comparing observed and calcu 
lated intensities (the latter uncorrected 
for the factors varying slowly with G). It 
is therefore concluded that 7 , which during 



FIG. i. POWDER PHOTOGRAMS OF ALLOY OF 20.2 PER CENT SILVER IN ALUMINUM. 

a, spectrum of 8 solid solution. As quenched from solution heat-treatment. 

b, spectra of 5 + 7 - Quenched and aged $ minutes at 387C. 

c, spectra of 6 + 7. Solution heat-treatment followed by cold-work; aged 16 hours at i75C. 



small cylindrical camera 2 in. in diameter 
and in some cases were duplicated in a 
4J^-in. camera. Direct radiation from a 
cobalt target operating at 37,000 volts 
was used. The specimens were rotated 
during exposure, which averaged one hour. 

Results 

The powder photograms showed that in 
addition to the equilibrium phase 7 there 
is a transition lattice 7 . 14 Typical diffrac 
tion spectra are listed in Fig. i. In Fig. la 
the diffraction lines of the 8 solid solution 
are alone present; in Fig. ib the transition 
lattice y f lines appear in addition to the 
8 lines; and in Fig. ic the 7 lines appear 
in addition to the ft lines. The Ka lines 
from the 7 structure that did not overlap 
lines of or 7 are listed in Table i ; it will 
be seen that their sin 2 values are fully 
accounted for by the quadratic form for the 
hexagonal system. The intensities are 
adequately accounted for by assuming a 
random distribution of atoms upon a close- 
packed hexagonal structure lattice, as will 



precipitation precedes the 7 phase, is a 
hexagonal close-packed solid solution (dis- 

TABLE i. Powder-diffraction Lines from 
y Lattice 





Intensity 


Sin 2 


Indices 


Observed 


Calcu 
lated 


Observed 


Calcu 
lated 


10. O 


M 


4 


o. 131 


o. 1302 


ro. i 


S 


18 


o. 166 


o. 1678 


10.2 


M 


6 


0.280 


0.2805 


10.3 


M 


18 


0.469 


0.4684 


20.0 


W 


3 


0.520 


0.5207 


20. I 


S 


18 


0-559 


0.5565 


00-4 


W 


4 


0-599 


0.6014 


20.2 


M 


6 


0.673 


0.6711 


10-4 


M 


6 


0-733 


0.7316 


20.3 


M 


18 


0.859 


0.8590 


12. O 


M 


6 


0.911 


0.9113 


12. I 


VS 


36 


0.949 


0.9489 



Sin 2 9 = 0.13018 O 
CoK a = 1.7866 oo 



4- hk + & 2 ) + 0.03759*- 
2.858 Co = 4.607 



ordered) with a = 2.858, C Q = 4.607, and 
CQ/CLQ = 1.612. The similarities between 
this lattice and the matrix on one hand 
and the equilibrium lattice 7 on the other 
hand provide an insight into the atomic 



138 PRECIPITATION FROM THE SOLID SOLUTION OF SILVER IN ALUMINUM 



movements during aging. These move- also contains streaks that must be dis- 

ments are simple and may be stated in tinguished from the aging streaks. The 

detail with considerable assurance of their streaks in Fig. 2h, from pure aluminum, 

reality, as we shall see below. are caused by thermal agitation of the 

TABLE 2. Aging of Annealed Wires and, Structures Present after Aging 



Aging 




























Tem 




























per 




























ature, 




























Deg. 




























C 




























160 


Time, hr. 


I 


4 


12 


24 


48 


96 
















Precipitate 












7 7 














173. ... 


Time, hr. 


I 


21 


47 


72 


120 


192 














228 


Precipitate 
Time, hr. 


1 A 


iMa 


y 

2 


y 
6 


7 
12 


7 X 


27 


48 


96 


120 


168 


240 


302. . . . 


Precipitate 
Time, hr. 


K 2 


$i 


t 


iH 


7 

3 


7 
12 


24 


48 


96 


7 


y f + 7(?) 


y 




Precipitate 


y 


y 


y 


y 


y 


7 


7 + 7 


7 + 7 


7+7 








387. . . . 


Time, Hr. 


Ms 


1 A 


1^2 




ej 


9 


25^ 


48 


I2O 








440 


Precipitate 
Time, hr. 
Precipitate 


$ 


4 

7+7 


I 

y + 7 


2 

y + 7 


7 + 7 
7 + 7 


7+7 
IO 
/ + 7 


7 + 7 


7 


7 









a */ and 7 in about equal amounts. 



The sequence of the appearance of 7 
and 7 was determined by the aging treat 
ments listed in Table 2. All aging treat 
ments from 1 60 to 440 C. first brought 
out the spectrum of 7 and later 7, the 
latter increasing in intensity at the expense 
of j f and indicating that 7 is formed by a 
transformation of 7 into the stable phase 
7. This behavior is similar to that shown 
by the & and phases in aluminum-copper 
alloys. 4 ~* 8 

Laue Photograms 

The size, shape, and orientation of the 
precipitating particles in the early stages 
of aging are conveniently shown by Laue 
photograms. These were made with the 
radiation from a cobalt target at 37,000 
volts, with exposures varying from 2 to 
12 hr. The analysis of the streaks and 
spots in the Laue patterns has been 
presented in detail in a previous paper 15 
and will be mentioned only briefly here. 
Typical patterns are reproduced in Figs. ia 
to 2h. In all Laue photograms shown here, 
the X-ray beam penetrated the crystal 
along [ooi] with [100] vertical; other 
orientations were analyzed but need not 
be reproduced. It is necessary first to show 
the pattern for pure aluminum, since it 



atoms 16 " 18 and are always radial. The 
streaks from the aluminum-silver alloy 
during aging are not radial, but follow 
ellipses (see, for example, Figs. ia and 2d). 
By plotting the streaks on a stereographic 
projection, in the way Laue spots are 
commonly plotted, it has been found that 
the streaks correspond accurately to great 
circles passing through the four { 1 1 1 } 
poles, and from a consideration of the 
reciprocal lattice or the Fourier transform, 
it is clear that this means that the lattice 
regions giving rise to the streaks have a 
relaxed third Laue condition; i.e., these 
regions are thin plates, responding to 
X-rays approximately as crossed gratings, 
and that these regions are therefore 
Guinier-Preston aggregates, lying upon the 
1 111} matrix planes. Similar diffraction 
effects have been observed in aluminum- 
copper alloys, 6 - 7 as mentioned above, but 
there is an important crystallographic 
difference between the two cases: in 
aluminum-silver alloys the thin plates are 
on the {111} planes, whereas in aluminum- 
copper the gratings are on {roo} planes. 

From the diffuseness of the streaks and 
the intensity distribution along them, the 
approximate thickness and diameter of 
the Guinier-Preston aggregates can be 



C, S. BARRETT, A. H. GEISLER AND R. F. MEHL 



139 



estimated, using the formulas for line 
widening from the particle size effect. The 
dimensions corresponding to the patterns 
in Fig. 2 are listed in Table 3 in the order 

TABLE 3. Approximate Dimensions in A. 
of Precipitate Corresponding to the 

Laue Patterns of Fig. 2 
LISTED IN ORDER OP INCREASING SIZE 



Fig. 


Thickness 


Diameter 


2b 


<IO 


<IOO 


2C 


<IO 


<IOO 


20, 


<IO 


<IOO 


2d 


<IO 


>IOO 


2e 


10 tO 100 


>IOO 


2/ 


10 tO 100 


>IOO 


2g 


>IOO 


>IOO 



of increasing size. At some stages of aging 
it is clear that there is a considerable 
range of particle size, but Table 3 lists 
only the larger ones. In Fig. 20 the streaks 
are so diffuse as to be invisible except 
where they overlap with other streaks. In 
this state the Guinier-Preston aggregate 
has a diameter of less than looA. and a 
thickness of a very few atom layers. Aging 



manstatten structure in this system. 2 By 
this time the precipitate has grown to a 
thickness sufficient to afford three-dimen 
sional diffraction, and the particles are 
io~ 6 cm. or more in all dimensions. Figs. ie 
and 2/ are somewhat earlier stages in the 
growth, where the thickness is roughly 10 
to i oo A. All of the Laue patterns obtained 
can be classified among the types illus 
trated in these figures. A summary of the 
aging treatments given one of the crystals 
studied is provided in Table 4, with the 
corresponding classification of the pattern 
after each treatment. The crystal was first 
given a solution heat-treatment of 16 hr. at 
556C., then quenched into a mixture of 
dry ice and acetone. Fig. 2a was obtained 
from the specimen in this state, indicating 
that the quench had not entirely sup 
pressed the formation of precipitate. After 
the 20 aging experiments listed in the 
table, the crystal was rehomogenized for 
15 hr. at 530 C., and quenched in ice 
water, and aged at i5oC.; subsequent 
aging treatments followed a similar sched- 



TABLE 4. Aging Treatments Yielding Types of Patterns of Fig. 2 and Interpretation in 

Table 3 



Aging Temper 
ature, Deg. 
C. 

20 


Time, days 
Type of pattern 
Time, hr. 
Type of pattern 
Time, hr. 
Type of pattern 
Time, hr. 
Type of pattern 
Time, hr. 
Type of pattern 





7 


22 


SO 


150 


280 

2C 

32 


40 


47 


59 


66 

2C 


82 
2C 


104 
2d 


150 



20. 

20. 

20 

20 


H2 

2b 

X* 

2b 

iH 

2g 
I 
2g 


2 


ii 


26 


200 


% 


2 


r 

2b 


*x 

26 


4 






320 


400 





produces streaks of progressively increas 
ing sharpness, and when the patterns of 
type id are obtained, there is no longer any 
appreciable widening arising from a limited 
diameter of the precipitate. When the aging 
is continued for about i hr. at 32oC., or 
for shorter times at higher temperatures, 
new spots begin to appear along the Laue 
streaks (as in Fig. 2g) at positions de 
manded by the orientation relationships 
derived in an earlier study of the Wid- 



ule. Air-cooling the specimen produced 
the type of pattern shown in Fig. 2/. Ex 
periments with other crystals will not be 
reported in detail, for they were less com 
plete and afforded no additional informa 
tion except showing that the various stages 
are passed through at different rates in 
different crystals. 

The times in Table 4 are not accurately 
comparable with the times in Table 2, for 
the latter were derived from data on poly- 



140 PRECIPITATION FROM THE SOLID SOLUTION OF SILVER IN ALUMINUM 



crystalline wires. A further limitation to 
the time scales in the tables is that the 
appearance of the X-ray films is somewhat 
dependent on the amount of exposure 



very thin plates of 7 . Whatever the ar 
rangement of atoms in the cluster, the 
fact that it forms on the (in) is good 
evidence that the lattice movements that 





I I 




FIG. 2. LAUE PHOTOGRAMS OF A SINGLE-CRYSTAL 
a, as quenched from solution heat-treatment. 
bj aged 5 minutes at 200 C. 

c, aged 50 days at room temperature. 

d, aged 104 hours at isoC. 



given: Type 20 can be made to look like 
2 b by underexposure. 

Despite the uncertainties in comparing 
the single- crystal and polycrystalline ex- 
periments, it can be definitely stated that 
the origin of the streaks in the Laue photo- 
grams is not the equilibrium phase 7, but is 
either platelike clusters of silver atoms on 
the (m) planes of the matrix, or simply 



ultimately form 7 have already begun. In 
the lack of evidence to the contrary, we 
strongly incline to the view that the clusters 
are in fact very small particles of the 
transition lattice 7 .* 



* A similar suggestion has been made above 
with respect to the similar set of circumstances 
in the aluminum-copper system. 



C. S. BARRETT, A. H. GEISLER AND R. F. MEHL 



141 



It happens that the axial ratios of 7 and shows that the 7 phase alone forms a well- 

7 differ so little that they cannot be dis- developed Widmanstatten figure (no 7 has 

tinguished on Laue photograms; but the yet formed in this specimen); a count of 

sequence of changes given in Table 4 the number of different trace directions in 



I f 



* 








I 

i 



X-RAY BEAM ALONG [ooi] WITH [loo] VERTICAL. 

e, aged 2}^ hours at 2ooC. 

/, aged 3 hours at 200 C. 

g, aged i hour at 32oC. 

h, 99.97 per cent aluminum annealed n hours at 5ooC and quenched. 



clearly indicates that the sharp spots, when 
they first appear in the Laue photograms, 
are 7 instead of 7. 

Micro structure 

Photomicrographs of some of the poly- 
crystalline wires used for the X-ray photo- 
grams are reproduced in Figs. 3, 4, and 5. 
Fig. 3, of a specimen aged 5 min. at 387C., 



each of several grains confirms the expecta 
tion that the plates of 7 are on (i 1 1) planes 
of the matrix. Another specimen aged for 
30 min. at the same temperature exhibits 
unusually long markings; apparently a 
number of parallel plates have grown 
laterally until they extend entirely across 
grains of the matrix. The precipitate is 
again wholly 7 . Fig. 5, of a specimen aged 



142 PRECIPITATION FROM THE SOLID SOLUTION OF SILVER IN ALUMINUM 



25^ hr. at the same temperature, shows 
plates again in a Widmanstatten pattern. 
X-rays show that this specimen contains 
7 and 7 in approximately equal amounts. 
Direction counts of a number of these 
grains again confirm (in) precipitation.* 
Samples containing 7 were difficult to 
etch; the microstructures observed (Figs. 3 
and 4) are somewhat diffuse and the edges 
of the 7 plate not clearly resolved. This 
was observed even when the 7 plates are 
relatively large. Plates of 7, however, were 
readily resolved, possessing well-defined 
sharp edges. Gayler 19 made similar observa 
tions on aluminum-copper alloys; she 
experienced difficulty in developing a sharp 
edge along one of the sides of rectangular 
6 plates, and took this as separate evidence 
of a coherency between 8 and the matrix. 
Such evidence, resting upon minor varia 
tions in etching attack and in arrangement 
and orientation of very small particles, is 
rather uncertain. 

Orientation Relationships 
It has been mentioned above that the 
first nuclei to give diffraction effects, the 
Guinier-Preston aggregates, are thin plates 
lying parallel to (m) planes of the matrix. 
It is shown by the Laue streaks that close- 
packed rows of atoms in the plane of these 
nuclei are parallel to the close-packed rows 
in the (in) matrix planes against which 
they lie.f 

At a later stage, when the 7 particles 

* From the approximate rates of the reaction 
7 to 7 established in the present study, it 
follows that the earlier Widmanstatten paper 
was concerned largely with the equilibrium 
phase, though some of the specimens may have 
contained both 7 and 7, but since the two 
lattices are indistinguishable in habit and only 
slightly different in structure, the crystallo- 
graphic statements of the earlier paper need no 
amendment. 

f This is most readily seen in diagrams of the 
reciprocal lattice 15 from the fact that each 
streak of the Laue photograms corresponds in 
reciprocal space to a line passing through the 
corresponding points of the reciprocal lattice of 
the matrix. The development of precipitate 
spots reproduced in Figs. 2e, 2/, and 2g, 
together with the analysis of 2g 16 also confirm 
this. 



have grown large enough to diffract as 
three-dimensional gratings, the same orien 
tation relationship holds, as can be deduced 
from the position of the diffuse precipitate 
spots in Figs. 2e and 2/ 15 and the sharp 
precipitate spots in Fig. 2g. A further proof 
is provided by the portion of a powder- 
diffraction pattern enlarged and reproduced 
in Fig. 6. In this, two rows of spots are 
shown close to each other, with each intense 
spot of the stronger row matched by an 
identical spot just opposite it in the weaker 
row. The stronger row consists of (in) 
spots from the matrix; the weaker row 
consists of (oooi) spots from the precipi 
tate, which is known to be 7 by its full 
spectrum on the same film. The photogram 
was prepared by rotating the polycrystal- 
line wire and therefore is a rotating crystal 
pattern of a number of differently oriented 
grains; from the geometry of such patterns 
it can be shown that only with the orienta 
tion relationships as stated above could 
there be the one-to-one correspondence be 
tween the two rows of spots seen in Fig. 6. 
A similar pairing of spots also appears in the 
second-order lines from these same planes. 
When 7 decomposes to 7, there is no 
reason to expect an alteration of orienta 
tion, for the former can be regarded as the 
latter under elastic strain, as we shall show 
later. The orientation is retained in rotating 
polycrystalline wire photographs similar 
to Fig. 6, which show (in) and (oooi) 
spots opposite each other (in both first and 
second orders). More direct proof is af 
forded by Laue photographs of single 
crystals that have been aged at high tem 
peratures until there could be no question 
but that all the precipitate is 7. These give 
patterns indistinguishable from Fig. 2g, 
proving the orientation mentioned (see 
reference to Fig. ig above). 

Discussion of Results 

There can be no doubt that there is com 
plete registry between the basal plane of 
7 and the (in) plane of the matrix on 
which it forms, for the atomic array on 



C. S. BARRETT, A. H. GEISLER AND R. F. MEHL 



these two planes is identical in pattern, 
interatomic distances, and orientation. 
Interatomic spacings that do not lie in this 



of planes of the face-centered lattice, 
ABCABC . . . , is altered into the se 
quence for the hexagonal close-packed 



plane are altered, however; the (in) plane lattice, ABABAB . . . These movements 




FlG. 3. POLYCRYSTALLINE WIRE AGED FlG. 4. POLYCRYSTALLINE WIRE AGED 30 

5 MINUTES AT 387C. X 2500. PRECIPITATE MINUTES AT 387C. X 2500. PRECIPITATE 
is 7 . is 7 . 

Etched in 0.5 per cent HF. 



spacing of 5, </<m), contracts 1.4 per cent 
in forming the (oooi) planes of 7 . On 
forming 7 from 7 there is a further con 
traction of 0.74 per cent in the (oooi) plane 
spacing, rf<oooi)> a&d an expansion of 0.73 
per cent in all atomic distances lying in 
this plane. These facts provide a basis for a 
consistent theory of the sequence of lattice 
changes. 

At those regions in the matrix where 
statistical fluctuations have produced a 
sufficiently high silver content,* parallel 
(in) planes of the matrix move over one 
another during precipitation in such a way 
as to convert the face-centered cubic 
lattice into the hexagonal close-packed 
lattice. This movement is in the direction 
[lolo]. If shifts of this kind take place 
with proper regularity, the sequence 

* See discussion of concentration fluctuations 
given in ref. i. 



are illustrated in Fig. 7. Fig. 70 shows the 
face-centered cubic 8 lattice with (in) 
as the base plane; in forming the 7 lattice, 
shown in Fig. 76, the upper two atom 
layers of the four shown move to the left, in 
the direction indicated by the arrows, 
taking positions at the tips of the arrows; 
the extent of translation is the same for 
these two layers. This produces the 7 
lattice shown in Fig. 76, as comparison of 
these two figures will show. The horizontal 
atom spacing does not change in the forma 
tion of 7 , but the perpendicular spacing 
decreases somewhat (1.4 per cent). In 
the formation of 7, shown in Fig. jc, there 
is no translation of atom layers, but merely 
a readjustment of all atom spacings.* 

* This crystallographic mechanism is identi 
cal with that proposed earlier for the same 
alloy 2 except that the earlier description did not 
include a consideration of the intermediate j 
step, which was then unknown; the mechanism 



144 PRECIPITATION EROM THE SOLID SOLUTION OE SILVER IN ALUMINUM 



It has been suggested 15 that the shifting 
of (in) layers of atoms necessary to create 
Y might have occurred in an irregular 
fashion to yield the Laue streaks observed. 




FlG. 5. POLYCRYSTALLINE WIRE AGED 
25*2 HOURS AT 387C. X 2000. PRECIPITATE 
IS 7 + 7- 

Etched in 0.5 per cent HF. 

For example, the plane sequence of the 
matrix ABC ABC . . . may change to 
the sequence of y A BAB A B . . . through 
an intermediate stage of staggered irregu 
lar sequence, e.g. ABABCABABABC 
. . . But such a lattice would in fact be 
merely matrix +7 in a lamellar arrange 
ment. The postulated shifting can thus 
create only 7 from the matrix there 
appears to be no possibility of the forma 
tion of an intermediate lattice between 

proposed introduced the idea of shearing move 
ments into considerations of phase changes of 

this type. 



the matrix and 7 . This reasoning supports 
the assumption that the Guinier-Preston 
aggregate is only 7 in a fine state of 
subdivision. 




FIG. 6. ENLARGED PORTION OF POWDEI 

PATTERN CONTAINING 5+7 SPECTRA. 
Spots on (111)5 ring paired with weake] 
spots on (0001)7 showing parallelism of thes< 
planes. Specimen aged ij^ hours at 387C. 

It is obvious that the matrix 5 musl 
strive to form the 7 immediately. The 
shearing movements of atom layers, whicl 
is the major lattice alteration necessar} 
for the conversion of d to 7, operates 
immediately, but the atoms on the (in 
matrix plane cohere to the atoms opposite 
on the (oooi) 7 , preventing a change ir 
their spacing until during growth th< 
strain which this entails becomes intoler 
able. At the 7 stage, therefore, the pre 
cipitate lattice has not broken away fron 
the matrix, but strains the matrix and ii 
itself strained by the matrix. The con 
traction of 1.4 per cent normal to th< 
(oooi) plane can probably be accom 



C. S. BARRETT, A. H. GEISLER AND R. P. MEHL 



modated by the matrix lattice without 
the formation of a crack at the interface, 
because of the thin platelike shape of the 
precipitate. 10 " 12 With increasing thickness 



Fig. 2c, were observed after aging at and 
below i5oC. but not after aging above 
this temperature. These streaks are charac 
teristic of an extremely small particle 





a, structure of 5 solid solution. 



} 2.6 7 S A *j 

b, structure of transition c, structure of equilibrium 

phase 7 . precipitate phase 7. 

FIG. 7. CRYSTAL LATTICE OF PHASES PRESENT IN AGED ALLOYS. 



of the 7 plates and the consequent increas 
ing strain, the tendency to break away 
increases and finally prevails, permitting 
the precipitate to take the lattice dimen 
sions of 7. 

From the appearance of the high order 
diffraction lines of 7, it may be concluded 
that the transition from 7 to 7 is a sudden 
one without intermediate steps; when the 
disregistry once begins, the precipitate 
probably snaps all the way to its final 
dimensions.* 

Nucleation theory requires that the 
smallest stable nuclei of the precipitate 
increase in size with increasing aging 
temperature. There is support for this 
proposed variation of nuclei size with aging 
temperature in the evidence presented in 
Fig. 2 and Table 4. The most diffuse 
streaks, those which give the patterns of 

* The transition from 7 to 7 is not unlike 
slip, for it involves displacements on slip planes. 
From this similarity, it seems reasonable to 
expect an instantaneous jump from 7 to 7 
when the stresses at the interface, aided by 
thermal agitation, start the displacement, for 
when any point on the interface breaks out of 
registry, there is a situation similar to that 
postulated in the recent theories of slip, in 
which slip resistance is thought to be greatly 
reduced by disregistry ("dislocation" ) on the 
slip plane. 



size. The streaks obtained by aging above 
i5oC. are sharper and are therefore 
evidence of larger particles of 7 . The 
quenching, though extremely severe, was 
unable to suppress all nucleation, for 
patterns like Fig. 20, were always obtained 
from the quenched alloy. It is interesting 
that in most of the experiments of Table 4 
the streaks of Fig. 20, were observed to first 
become much weaker, as in Fig. 2b before 
the intense sharp streaks of the later stages 
of aging appeared. This suggests that the 
particles present after quenching may have 
redissolved in part before a new set was 
precipitated. 

If 7 is merely 7 held by elastic strains 
from assuming the normal lattice spacings 
of 7, it seems justified to calculate the 
magnitude of the stresses in 7 from the 
known values of the strains. The elastic 
constants of the 7 phase are unknown, but 
a rough idea of the magnitude of the 
stresses can be obtained by assuming the 
elastic constants for either aluminum or 
silver (which do not differ greatly) and 
applying the elastic theory equations for 
an isotropic medium. Let the stresses be 
referred to two orthogonal axes, X and Y, 
in the basal plane and a third, Z, along 



146 PRECIPITATION FROM THE SOLID SOLUTION OF SILVER IN ALUMINUM 



the c axis of the crystal; the three principal 
stresses cr x , <r y , <r s are related to the three 
principal strains e x , e }/ , 6 Z by the three 
simultaneous equations 



Now *, and z may be evaluated from the 
difference between the lattice dimensions 
of 7 (strained) and 7 (unstrained) in the 
X, Y, and Z directions and have the fol 
lowing values: x = e u (2.858 2.870)7 
2.879 = 0.0073; e z = (4-607 - 4-573)/ 
4-573 = 0.0074. Inserting these values in 
the equations and assuming the values of 
E = 10,300,000, v = 0.34 as in aluminum 
alloys (the corresponding values for silver 
would not alter the calculations signifi 
cantly) the following stresses are obtained 
in pounds per square inch: 

o> = (T,, = 115,000 
<TZ = i, 800 

where the negative sign indicates com 
pression. The value 1,800 for <r z is, of 
course, less than the error of measurement 
and calculation and could as well be con 
sidered zero, permitting the assumption 
that all stresses exerted by the matrix on 
the precipitate are directed parallel to the 
basal plane as if the precipitate is glued 
to the matrix (in) plane and that there 
are negligible stresses normal to this plane 
(the c-axis contraction resulting from the 
basal plane stresses only). It might be 
pointed out that the matrix is unable to 
exert great stresses in the direction of the 
c axis of the precipitate, because of the 
piatelike shape of the particles, 10 " 12 which 
may explain the low value. It is also con 
ceivable that there is an actual crack along 
one face of the precipitated plate, where 
the plate has been drawn away from the 
matrix by the contraction resulting from 
the formation of 7 , but this is unlikely. The 
contraction is so slight that it would 



amount to as much as half of the inter- 
planar spacing d (n \} only if the plates were 
about 100 A. thick or more; that is, in a 
late stage of the process. The strain in the 
matrix around the edges of the 7 plates 
seems appropriate to facilitate further 
growth of the plate edgewise, for it must 
be contracted toward the spacing J ( ooon 
of 7 , and may account for the extended 
plates of Fig. 4. 

The sequence of events discussed above 
are in entire accord with the expecta 
tions outlined in earlier papers 1 and 
with Nabarro s theoretical treatment of 
stresses. 10 " 12 Assuming that the breaking 
away of a precipitated particle from the 
matrix plane on which it forms requires 
roughly the same energy as the melting 
of a monatomic layer of metal at the 
surface of the particle, Nabarro calculates 
that the particle will break away when its 
thickness equals 



where c\\ and c\i are the elastic moduli, A 
the latent heat of melting, / is the mechan 
ical equivalent of heat, and 8 is the misfit 
between the two lattices in the matched 
plane. If the constants for silver are in 
serted, the thickness for breaking away is 
roughly 107 X io~ 4 / 2 and for the case 
that Nabarro considered, the precipitation 
of silver from copper, breaking away will 
occur when the particle is only two atoms 
thick. But in the present instance, where 
the misfit between the unstrained (in) 
and (oooi) planes is only (2.879 ~~ 2.858)7 
2.858 = 0.0073, the same constants would 
predict breaking away at about 320 atom 
layers, or 73oA. The theory is thus in 
accord with the fact that sharp Debye 
rings can be obtained from 7 before it 
becomes large enough to break away and 
become 7. 

The stresses in the matrix that keep the 
precipitate in the strained 7 condition are 
beyond much doubt an important con 
tributor to the hardening. 20 It is an old 



C. S. BARRETT, A. H. GEISLER AND R. F. MEHL 



147 



idea that strains are set up during aging 
and that these are important in producing 
age-hardening. The theories were based 
on the volume changes accompanying 
precipitation and more vaguely on pre 
cipitate matrix interaction, but in recent 
years 1 - 4 " 8 - 20 the importance of the effect of 
registry on strain has been pointed out. In 
the aluminum- silver system considered 
here, there is a decrease in volume on the 
formation of 7 from the 5 solid solution, 
from 16.50 to i6.4iA. 3 per atom; thus 
broken-away 7 does not fit perfectly into 
its space in the 5 lattice and strain must 
result, but this must exert a minor harden 
ing effect, for it is observed that the 
formation of 7 from 7 in the aluminum- 
silver system and of from 6 f in the 
aluminum-copper system are accompanied 
by softening. Evidently at least in these 
systems a different origin of strain must be 
sought. 

With the demonstration of orientation 
relationships between precipitate and 
matrix, 2 and with the correlated near- 
matching of atom patterns and spacings 
at the interface and the strains implied by 
this imperfect registry, a basis is available 
for a more detailed study of age-hardening 
strains. When transition lattices occur, as 
they do in the aluminum-copper and 
aluminum-silver systems, the state of 
strain is more clearly apparent and the 
degree of strain is susceptible to calculation. 
Such strains obviously must interfere with 
slip and thus harden; they may also be at 
least in part responsible for the "anoma 
lies 3 occurring in property changes during 
age-hardening. 1 Further progress in the 
theory of age-hardening will be facilitated 
by a more detailed understanding of the 
intensity and distribution of these strains 
in the matrix. It appears possible to make 
approximate calculations of these in the 
present instance with elastic theory form 
ulas and this is being attempted. 

The Guinier-Preston aggregate, the 
transition lattice, and the equilibrium 



precipitate lattice in the aluminum-silver 
system are similarly arranged geometri 
cally. In the aluminum-copper system the 
same geometrical similarity obtains among 
the same stages, but the geometrical 
arrangement itself differs from that for 
aluminum-silver alloys. In each case the 
geometry of the Widmanstatten figure 
formed by the equilibrium precipitate 
appears also in the earlier stages. It appears 
therefore that the several lattice stages 
accompanying precipitation from solid 
solution may be regarded simply as succes 
sive steps in a single process of lattice 
alteration, which converts the matrix 
lattice to the lattice of the equilibrium 
phase. 

SUMMARY 

1. The lattice changes that accompany 
precipitation from the solid solution of 
silver in aluminum (20 per cent Ag) have 
been studied. 

2. This alloy forms Guinier-Preston 
aggregates, platelike in shape, upon the 
(in) matrix plane. These aggregates are 
small at low aging temperatures and larger 
at high aging temperatures; they grow 
with aging times. Their approximate sizes 
have been determined. It is possible that 
the aggregates are merely the transition 
lattice 7 in a high degree of dispersion. 

3. The alloy forms a transition lattice, 
7 , that is close-packed hexagonal. It also 
is platelike in shape and deposited on the 
(in) matrix plane. The spacing in 7 on 
the (oooi) 7 plane is identical with the atom 
spacing on the (in) matrix plane to which 
it lies parallel; the lattices are therefore 
coherent at this interface. Since the 7 
lattice differs from the equilibrium pre 
cipitate lattice 7 only in interatomic 
distance, 7 may be regarded as 7 strained 
from its ordinary dimensions by the lattice 
coherency between it and the matrix. The 
magnitude of this strain has been cal 
culated, and the stresses that cause it have 
been computed. 



148 PRECIPITATION FROM THE SOLID SOLUTION OF SILVER IN ALUMINUM 



4. The lattice movements that trans 
form the matrix to the equilibrium pre 
cipitate lattice have been inferred from the 
orientation relationships observed. 

5. Since the Gumier-Preston aggregate, 
the transition lattice y and the equilibrium 
lattice are all arranged in a similar way 
geometrically, it may be concluded that 
these stages in precipitation are merely suc 
cessive steps in a single process of lattice 
alteration, which converts the matrix lat 
tice to the lattice of the equilibrium phase. 

ACKNOWLEDGMENT 

The authors wish to express their thanks 
to the officials of the Aluminum Company 
of America, who made this work possible 
through the grant of a fellowship to one 
of them, and also to other members of that 
company for many profitable discussions. 

REFERENCES 

1. R, F. Mehl and L. K. Jetter: The Mechanism of 

Precipitation from Solid Solution. The Theory 
of Age-hardening. Amer. Soc. Metals, Harden- 
ability Symposium, October 1930. 

2. R. F. Mehl and C. S. Barrett: Studies upon the 

Widmansttten Structure, I Introduction. 
The Aluminum-silver System and the Copper- 
silicon System. Trans. A.I.M.B. (1931) 78, and 
later papers published by the Institute. 

3. R. F. Mehl, C. S. Barrett and_ P. N. Rhines: 

Studies upon the Widmanstatten Structure, 
III. The Aluminum-rich Alloys of Aluminum 
with Copper , and of Aluminum with Magne 
sium and Silicon. Trans. A.I.M.E. (1032) 99, 
203. 

4. G. Wassermann and J.^Weerts: On the Mechan 

ism of the Precipitation of CuAh in a Harden- 
able Copper-aluminum Alloy. Metallicirtsckaft 
(1935) 14, 605. 

5. W. L. Fink and D. W. Smith: Age-hardening of 

Aluminum Alloys, I Aluminum-copper 
Alloys. Trans. A.I.M.E. (1936) 122, 284, 

6. J. Calvet, P. Jaquet and A. Guinier: The Age- 

hardening of a Copper-aluminum Alloy of 
Very High Purity. Monthly Jnl. Inst. Metals 
(April 1939) 177. 

7. G. D. Preston: The Diffraction of X-rays by an 

Age-hardening Alloy of Aluminum and Copper. 
The Structure of an Intermediate Phase. Phil. 
Mag. (November 1938) 855. 

8. G. D. Preston: The Diffraction of X-rays by Age- 

hardening Aluminum-copper Alloys. Proc. 
Roy. Soc. (1938) i67-A, 526. 

9. P. D. M erica: The Age-hardening of Metals. 

Trans. A.I.M.E. (1932) 99, 25. 

10. N. F. Mott and F. R. N. Nabarro: An Attempt to 

Estimate the Degree of Precipitation Hard 
ening with a Simple Model. Proc. Phy. Soc., 
(1940) 52, 86. 

11. F. R. N. Xabarro: The Influence of Elastic Strain 

on the Shape of Particles Segregating in an 
Alloy: Proc. Phys. Soc. (ip4o) 52, 90. 

12. F. R. N. Nabarro: The Strains Produced by Pre 

cipitation in. Alloys. Proc. Roy. Soc. (1940) 
I75-A, 519. 

13. M. fiansen: Der Aufbau der ZweistofHegierun- 

gen. Berlin, 1936. Julius Springer. 



14- L. Guillet and L. Guillet, Jr.: Sur le durcissement 
structural des alliages aluminium-argent riches 
en aluminium. Compt. rend. (iQ39) 209, 79. 

15. C. S. Barrett and A. H. Geisler: Atomic Distribu 

tion in Aluminum-silver Alloys during Aging. 
Jnl. Applied Physics (Nov. 1940) 11, 733. 

16. G. D. Preston: Diffraction of X-rays by Crystals 

at Elevated Temperatures. Proc. Roy. Soc. 

(1939) 172, 116. 

17. W. H. Zachariasen: Theoretical Study of the 

Diffuse Scattering of X-rays by Crystals. Phys. 
Rev. (1940) 57, 597- 

1 8. S, Siegle and W. H. Zachariasen: Preliminary 

Experimental Study of New Diffraction 
Maxima in X-ray Photographs. Phys. Rev. 

(1940) 57, 795- 

19. M. V. Gayler: The Aging of High- purity 4 per 

cent Cu-Al Alloy, Part II. Jnl. Inst. Metals 
(1940) 66, 72. 

20. W. L. Fink and D. W. Smith: Age-hardening of 

Aluminum Alloys, IV Discussion of the 
Theory. Trans. A.I.M.E. (1940) 137, 95. 

DISCUSSION 

(W. L. Fink presiding) 

D. W. SMITH,* New Kensington, Pa. I 
wish to congratulate the authors on this very 
excellent paper. Dr. Barrett, in particular, 
deserves our appreciation for his lucid explana 
tion of Guinier-Preston zones given in the oral 
presentation. 

The authors point out two possible mecha 
nisms of precipitation: (i) a three-stage process 
in which Guinier-Preston zones are formed in 
some distinct but unknown manner, then 
actual precipitation of a second phase having 
a transition lattice structure, and finally a 
transformation of the precipitate to the equilib 
rium lattice structure; and (2) a two-stage 
process, which makes no distinction between 
the formation of Guinier-Preston zones and 
the precipitation of a transition phase but 
which recognizes the transformation of the 
precipitate to the equilibrium structure. 

Let us examine the facts to establish which 
of the two mechanisms appears the more 
logical. In the first place, the X-ray method 
used to detect the formation of the so-called 
Guinier-Preston zones is not capable of 
distinguishing between an agglomeration of 
solute atoms on certain crystallographic planes 
of the solvent lattice in the form of thin 
platelets (which was that condition originally 
proposed by Guinier and Preston) and the 
actual precipitation of a second (transition 
lattice) phase in the same form. Furthermore, 
the present authors have shown a distinct 
difference in location of the Guinier-Preston 
zones depending on the alloy system in Al-Cu 



* Aluminum Research Laboratories. 



DISCUSSION 



149 



alloys the zones form parallel to the {001} 
planes of the aluminum-solid solution while in 
Al-Ag alloys they form parallel to the {in} 
planes of the aluminum-solid solution. Finally, 
Dr. Mehl and his collaborators have already 
shown, in their studies on Widmanstatten 
structures, that the precipitate phases form 
as plates parallel to the same planes to which 
the Guinier-Preston zones lie parallel in the 
respective parent solid solutions. 

In view of these facts, it seems logical that 
the two-stage precipitation mechanism should 
be given more emphasis than the three-stage 
process, because: If the Guinier-Preston zones 
were merely local regions of high concentration 
of either copper or silver atoms in the parent 
solid solution, it would be expected that their 
location would be controlled only by the 
lattice type of the parent solid solution; on the 
other hand, if they were small precipitate 
platelets, they would be expected to form 
according to some orientation relationship 
(Widmanstatten) between the parent and 
precipitate lattices, which is the true case. 

L. W. KEMPF,* Cleveland, Ohio. Specula 
tion regarding the proper nomenclature for 
the Guinier-Preston phenomena is interesting 
and useful. It appears that the term precipita 
tion by general consent and usage will be 
utilized in referring to all the events accom 
panying the disintegration of supersaturated 
solid solutions in spite of the much more 
limited meaning assigned to the term by the 
general theory of heterogeneous equilibria. 

Perhaps an even more interesting field for 
speculation is the correlation between structure 
and mechanical properties. Nothing appears in 
the paper regarding hardness or other mechan 
ical properties. On the basis of data gathered 
some time ago on alloys of this type, I would 
expect maximum hardness to be reached with 
not more than about 20 hr. aging at i5oC. 
or not more than 2 or 3 hr. at 2ooC. As is usual 
in such precipitation-hardening systems, the 
hardness will probably be higher the lower the 
aging temperature. If approximately these 
relationships hold for the alloy investigated by 
the authors, it would appear that maximum 
hardness and presumably maximum strength 
are achieved prior to the appearance of appreci 
able quantities of y f or 7. In other words, in 

* Metallurgist, Aluminum Research Laboratories. 



this system, and perhaps also in the aluminum- 
copper system, the formation of Guinier- 
Preston zones is the most effective hardening 
process; and the development of appreciable 
quantities of 7 and 7 from these zones results 
in softening. I wonder whether the authors 
have any data that might be used to establish 
a correlation between the structure and prop 
erties of the alloy they investigated? 

C. S. BARRETT AND R. F. MEHL (authors 
reply). Dr. Smith is clearly in agreement with 
the final paragraphs of the Discussion of 
Results and the Summary but believes that 
more emphasis should be given to the theory 
that the Guinier-Preston aggregate is merely 
-y in a fine state of dispersion. This theory 
was the keynote of the paper, it will be noted, 
and the authors went as far as they thought 
was justified when the paper was written. 
The experimental work has been continued on 
these and other alloys and it is now possible to 
report additional support for the theory. 

During precipitation from aluminum-rich 
alloys of aluminum and magnesium, it is 
possible to observe streaks on Laue photograms 
that are very similar to those in aluminum- 
copper alloys. Analysis of these and of the 
Widmanstatten structure shows that the 
streaks are caused by thin plates forming on 
the (100) planes of the aluminum-rich matrix. 
The important factor in this experiment is the 
very close similarity in scattering power be 
tween the aluminum and magnesium atoms. 
The two kinds of atoms are so nearly alike in 
their ability to scatter X-rays that it would be 
impossible to detect mere segregation of 
magnesium atoms on the lattice of aluminum 
with the type of photogram we have used. 
Accordingly, we must ascribe the Laue streaks 
in this system to small platelike regions in 
which the structure is no longer than that of the 
matrix and has become that of the transition 
lattice, or perhaps has become an irregular 
structure that may be described as a mixture 
of the two or a compromise between the two. 
We have suggested and discussed an irregular 
or incomplete transition lattice in the paper of 
reference 15. This is the first case that has 
been investigated in which the segregation 
theory and the precipitation theory can be 
directly distinguished experimentally, and the 
fact that Laue streaks are found is more direct 



I$0 PRECIPITATION FROM THE SOLID SOLUTION OF SILVER IN ALUMINUM 



support for the precipitation theory than has 
been given by any of the previous observations. 
The reasoning advanced in the paper and 
reviewed by Dr. Smith leads to the conclusion 
that in other alloy systems also there is no 
distinction between the formation of what we 
have called Guinier-Preston aggregates and 
true precipitation. 

We agree with Dr, Kempf that the term 
"precipitation" may well be applied to all 
events accompanying the disintegration of the 
solid solution. Hardness measurements have 
now been made and correlated with powder 
diffraction spectra of 7 and 7 in the aluminum- 
silver system and it is definitely established 



that the transition lattice 7 alone can be seen 
in alloys aged to their maximum hardness. For 
example, with a 20 per cent silver alloy the 
maximum hardness with aging at 225C. is 
reached in about 3 hr.; 7 first appears in 
the powder patterns in one hour, whereas the 
stable precipitate 7 does not appear until the 
alloy is very much overaged (more than 40 days 
in one experiment). Similar results have been 
obtained with 10 and 30 per cent Ag alloys. 
It is safe to conclude that the hardening is not 
related to the stable phase in this system, but 
probably is associated with the stresses sur 
rounding the particles of the transition lattice 
as has been pointed out in the paper. 



Precision X-ray Study of the High-silver Aluminum-silver Alloys 

BY FRANK FOOTE,* JUNIOR MEMBER, AND ERIC R. JETTE,! MEMBER A.I.M.E. 

(Cleveland Meeting, October 1940) 



IN recent years the constitution of the 
high-silver aluminum-silver alloys has been 
extensively investigated. Hansen 1 has re 
viewed the literature to 1935. More re 
cently, Obinata and Hagiya, 2 Hofmann and 
Volk, 3 Taziri 4 and Tishchenko 5 have con 
firmed the general features of the phase 
diagram in the region around 25 atomic per 
cent aluminum. 

In this paper we present the results of 
high-precision measurements of lattice con 
stants of the silver-rich alpha solid solution 
in the homogeneous range and the neigh 
boring two-phase regions. These results 
are used to establish the nature of the solid 
solution, the identity of directly measured 
densities with those calculated from lattice 
constants and the homogeneity limits of the 
alpha solid solution with respect to three 
different neighboring phases. A study of the 
earlier investigations indicated that the last 
two subjects in particular required new and 
accurate data. 

EXPERIMENTAL PROCEDURE 

Preparation of the Alloys. The alloys 
were made from high-purity silver and 
aluminum. The silver was "Proof Silver," 
99.999 per cent pure, supplied by the U. S. 
Metals Refining Co. The aluminum was 
99.97 per cent pure (contained Si, 0.009 P er 
cent; Fe, 0.0016 per cent; Cu, 0.004 per 



Manuscript received at the office of the Institute 
June 13, 1940. Issued as T.P. 1229 in METALS 
TECHNOLOGY, September 1940. m 

* Instructor in Metallurgy, Cooper Union, JNew 
York, N. Y. 

t Professor~of Metallurgy, School of Mines. Colum 
bia University, New York, N. Y. 

1 References are at the end of the paper. 



cent) and was supplied by the Aluminum 
Company of America. The lattice constants 
of these two materials .have been reported 
by Jette and Foote. 6 Weighed amounts of 
the two metals were placed in alundum 
crucibles and melted in vacuo in a high- 
frequency induction furnace. The melts 
were allowed to cool in the crucible, the 
ingots thus obtained weighed about 10 
grams. These ingots were then filed clean, 
hamered, again cleaned and given a homo 
genizing anneal in vacuo of from 6 to 15 
days at temperatures ranging from 500 to 
695C. The ingots were air-cooled from the 
annealing temperature. 

Lattice-constant Measurements. Care 
fully prepared filings for X-ray analysis 
were sealed into evacuated quartz or Pyrex 
tubes and annealed for various periods of 
time at suitable temperatures. The furnaces 
used were vertical, rapid-quenching fur 
naces, the samples being quenched by 
crushing the sample tube under water by 
means of a heavy plunger. 

The general X-ray technique has been 
previously described. 6 The cameras were of 
the symmetrical, back-reflection, focussing 
type. K a radiation from a cobalt-nickel 
alloy target was used. The wave-length 
values were taken from the second edition 
of Siegbaum s book. 7 A correction for 
the index of refraction was applied and the 
lattice constants were calculated by the 
method of Cohen. 8 - 6 The lattice constants 
were calculated to a common temperature 
of 25C., using the value of the linear expan 
sion coefficient of pure silver as 18.9 X io~ 6 . 
Presumably the expansion coefficient for 



152 



PRECISION X-RAY STUDY OF ALUMINUM-SILVER ALLOYS 



the high-silver solid solution is not far 
different from that of pure silver and in any 
case the correction is very small. 

Density Measurements. The density of 
three of the single-phase alloys was deter 
mined by the loss-of-weight method. The 
density liquid was mono-brom-benzene. 
The density of the brombenzene was cal 
culated from the equation given in Inter 
national Critical Tables (vol. 3, p. 29). 
This equation was checked by pycnometer 
weighings and found to be essentially cor 
rect. As a density liquid, brombenzene has 
a number of advantages. It is easily purified 
by simple distillation, has a high density 
of about 1.5 grams per cubic centimeter, 
moderate viscosity, low vapor pressure and 
low surface tension. The samples used were 
annealed filings. All measurements were 
corrected to 25C. using 3 X 18.9 X icr 6 
for the volume coefficient of expansion. In 
all cases, this correction was small. 

Chemical Analysis. All alloys were ana 
lyzed for aluminum by precipitating as 
Al(OH)s with ammonium hydroxide and 
weighing as A1 2 3 , closely following the 
method given by Hillebrand and Lundell. 9 
The method was checked by analyzing syn 
thetic mixtures containing known amounts 
of silver and aluminum. The analytical 
values were close to the values calculated 
from the weights of the pure metals used in 
making the alloys. Determinations were 
made on annealed filings used for the den 
sity and X-ray measurements and on mill 
ings from the annealed ingots. No significant 
differences between these values were found. 
All compositions were converted to atomic 
per cent, using for the atomic weights: Ag, 
107.88; Al, 26.97. 

Lattice Constants and Densities of the 
Single-phase Alloys. The lattice-constant 
data for the single-phase alloys have been 
tabulated in Table i and plotted in Fig. i. 
The heat-treatments refer to the final 
powder anneal. The standard errors of the 
the lattice-constant measurements average 
o.ooo 10 A CT . The data can be satisfactorily 



represented by the least-squares straight 
line: 

a = 4.07787 - 0.001253 (At % Al) [i] 

The average deviation of the individual 
measurements from this line is o.ooo 10 A cr . 

TABLE i. One-phase Alloys 



Alloy 


Heat-treatment 


Atomic 

P<ar (""pnf 


Lattice 




Hr. 


Deg. C. 


Jrer ^eiit 
Al 


Constant, 
ao (25C.) 


Ag 


(Ave. of 7) 


(ref. 6) 


0.00 


4.07787 


I 


2 


721 


4.00 


4.07288 


12 


116 


548 


10.76 


4.06452 


2 


2 


721 


11.22 


4.06382 


II 


146 


538 


14.24 


4.05980 


4 


2 


693 


15-75 


4-05794 


4 


40 


500 


15-75 


4-05793 


3 


2 


721 


17-35 


4.05626 


3 


40 


500 


17-35 


4.05608 


3 


1.5 


720 


17-35 


4-05592 


7 


112 


598 


17-45 


4.05621 


7 


5 


520 


17-45 


4.05589 


10 


Ii6 


548 


18.28 


4-05522 


10 


230 


598 


18.28 


4-05518 



Equation i can be solved for atomic per 
cent Al: 

At % Al = 3255-16 - 798.250(00) [2] 

The average deviation of the individual 
measurements from this line is 0.08 atomic 
per cent Al. The lattice constants show 
large negative deviations from Vegard s 
rule. 

Westgren and Bradley, 10 Ageew and 
Shoyket, 11 Barrett 12 and Phelps and 
Davey 13 have previously measured the lat 
tice constants of certain of these high-silver 
alloys. The results of the first three of these 
investigations were somewhat higher than 
those reported in this paper; the results of 
the fourth were considerably lower. West 
gren and Bradley measured the lattice con 
stants of two alloys within the solid solution 
range; their claimed precision was 0.002 A cr . 
Hume-Rothery pointed out in discussion 11 
that the results of Ageew and Shoyket were 
neither self -consistent nor properly handled. 
Barrett s results showed the quenching 
effect 14 obtained with samples of moderate 



FRANK FOOTE AND ERIC R. JETTE 



size. The results of Phelps and Davey 
are entirely out of line with other deter 
minations and cannot be considered as 
precision measurements. 

//.O 



in which n is the number of atoms per unit 
cell, M is the average atomic weight, ao is 
the lattice constant in " crystal Angstroms " 
(i.e., io 3 X -units), N is Avogadro s number 

4,06 



o - foote and Jeffe 

e - Borreff 

- Phe/ps ancf >cn/ey 




4.O7 



4.06 



4.05 



9.0 



404 



p IG> 



/O /5 

Atomic Per Cent A/uminum 
[ . LATTICE CONSTANTS AND DENSITIES OF SINGLE-PHASE ALLOYS. 



Phelps and Davey, 13 Barrett, 12 Kokubo 15 
and the present authors have measured the 
densities of a number of these high-silver 
solid solution alloys. The data are summa 
rized in Table 2 and plotted in Fig. i. With 
a few exceptions, the measured densities 
agree very well among themselves. The 
solid line of Fig. i is the theoretical density 
calculated on the basis of simple substitu 
tion of aluminum for silver atoms in the 
silver lattice. The formula used for this 
calculation was: 



[3] 



and /is the ratio between the absolute and 
the relative Siegbaum scale of X-ray wave 
lengths. Jette and Foote 16 have evaluated 
the factor K = io 24 //W from published 
data on calcite. The numerical value of K 
was found to be 1.65023 0.00021. The 
density formula thus becomes: 

p - 1.65023 X -^3 b l 

where: n = 4 

M = 107.88 - 0.8091 (at. % Al) 

(simple substitution) 
a = 4-077^7 - 0-001253 (at. % Al) 
(eq. i) 



1 54 



PRECISION X-RAY STUDY OF ALUMINUM-SILVER ALLOYS 



The last column of Table 2 gives the com 
parison between the measured and cal 
culated densities. For the most part, the 
agreement is excellent. In particular, the 
agreement with Barrett s measured densi- 

TABLE 2. Density of Single-phase Alloys 



Ob 
server* 


Atomic 
Per 
Cent 

Al 


Lattice 
Constant 
ao C25C.), 
Calcu 
lated 


P (2SC.) 


Ap X I0 3 


Calcu 
lated 


Meas 
ured 


FJ 


o.oo 


4.07787 


10.501 


10 492 


- 9 


B 


O. 00 


4.07787 


10. 501 


10.489 


12 


PD 


0.00 


4 07787 


10.501 


10,50 


i 


PD 


2.47 


4-07477 


10 330 


10.32 


10 


PD 


3.85 


4-07304 


10 234 


10 23 


- 4 


& 


4.00 

7.76 


4.07286 
4. 06814 


10.224 
9.961 


id 213 
9 966 


ii 
+ 5 


PD 


10.81 


4-06432 


9-747 


9 737 


10 


B 


11.04 


4. 06404 


9-731 


9.726 


- 5 


B 


11.04 


4. 06404 


9-731 


9-727 


- 4 


FJ 


II. 22 


4 06381 


9 7i8 


9-701 


-17 


PD 


14.25 


4.06001 


9.503 


9-507 


+ 4 


PD 


15.36 


4.05862 


9 424 


9-362 


-62 


K 


15-75 


4.05814 


9 397 


9-423 6 


4-26 


K 


15-75 


4 05814 


9 397 


9.36l c 


-36 


B 


17.48 


4-05597 


9 273 


9-274 


+ i 


B 


17.51 


4 05593 


9 271 


9-273 


+ 2 


PD 


17.51 


4-05593 


9-271 


9.216 


-55 


FJ, Foote and Jette; PD, Phelps and Davey; 


B, Barrett, K, Kokubo. 


6 Cast, cold-worked and annealed. 


Cast and annealed. 



ties is uniformly good and considerably 
better than the agreement with his own 
calculated densities. We may conclude that 
there is no significant difference between 
directly measured densities and those cal 
culated from X-ray data, provided both 
sets of measurements are carefully done and 
the proper formula used for the density 
calculation. This fact has not previously 
been fully established. Furthermore, in 
view of the close agreement between the 
measured and calculated densities, we are 
justified in believing that the solid solution 
of aluminum in silver is of the simple 
substitutional type 12 and that any other, 
more complex, theory of solid solution is 
unnecessary. 13 

Solid Solution Limit. Data on the two- 
phase alloys used in determining the alpha 
homogeneity limits are tabulated in Table 3 
and plotted in Fig. 2. The solid solubility 
limit was calculated by substituting the 



measured values of the lattice constants in 
eq. 2. Also, in Fig. 2, the liquidus as deter 
mined by Hoar and Rowntree 17 and the 
solidus as determined by Hume-Rothery, 
Mabbott and Evans 18 are plotted. 

In order to keep the alpha-phase lines 
strong and to avoid interference by the 
lines of the second phase (particularly by 
the gamma phase), it is desirable that the 
two-phase alloys be only slightly higher in 
aluminum than the homogeneity limit. At 
the same time, we must be sure that the 
alloys are actually two-phase. In doubtful 
cases, the samples were photographed in 
an identification camera (Phragmen type) 
to check the presence or absence of the 
second phase. 

TABLE 3. Two-phase Alloys 



Alloy 


Heat-treatment 


Lattice 
Constant, 
ao (2SC.) 


Solubility 
Limit 


Hr. 


Deg. C. 


5 


I 


748 


4.05485 


18.37 


5 


o.S 


737 


4 0549 


18,26 


5 


2-5 


698 


4 05425 


18 84 


5 


3 


693 


4 05425 


18 84 


14 


6 
6 


6 5 5 
655 


4-05337 
4-0534o 


19-55 
19-53 


14 


43 


617 


4-0525 


20 25 


14 


40 


602 


4 05233 


20.34 


14 


329 


574 


4 05243 


20. 26 


14 


341 


507 


4 05283 


19 99 


14 


490 


453 


4. 0527,- 


20.05 


13 


490 


453 


4-05302 


19 83 


5 


747 


430 


4.05460 


18 57 


5 


5 


412 


4 0557 3 


17 67 


10 


407 


401 


4 0564^ 


17. 10 


10 


744 


374 


4 osSiJ 


15-74 


10 

II 


453 
H73 


352 
306 


4-0599 4 | 
4 06155 


14-31 
13.06 


II 


5491 


263 


4-06383 


ii 16 


II 


1124 


198 


4.06715 


8.56 



1 Total compositions: 

ALLOY ATOMIC PER CENT AL 

5 20 5 

13 20. o 

14 21 5 

10 18.3 

11 14.2 



The solid solubility limit is not a con 
tinuous line, but consists of three distinct 
parts. The breaks at B and C correspond to 
changes in character of the second phase. 
The portion AB represents the solubility 
limit with beta, BC with gamma, and CD 
with beta prime (Ag 3 Al) as the adjoining 
phase. 



FRANK FOOTE AND ERIC R. JETTE 



155 



The point A (obtained by extrapolation mines the temperature at which the pen- 
to 779C., the peritectic temperature found tectoid reaction for the formation of Ag 3 Al 
by Hoar and Rowntree) is placed at 17.84 takes place. Point C is placed at 4S2C. 
atomic per cent Al. The break in the solu- and 19,92 atomic per cent AL This tem- 




/OO 



Atomic 

> 2. PHASE DIAGRAM OF HIGH-SILVER ALUMINUM-SILVER ALLOYS. 



bility limit at B determines the eutectoid 
temperature for the decomposition of the 
beta phase. This point is placed at 6o9C. 
and 20.34 atomic per cent Al. The eutectoid 
temperature has been previously reported 
at 6i5C., 2 6ioC., 11 19 6o9C., 20 6o6 C., 5 21 
and 570 to 6osC. 3 The break at C deter- 



perature has been previously reported as 
4 5 6C., 5 42oC., 2 4 oo C., n and 370 to 
465C. 3 It is of considerable interest that 
the eutectoid and peritectoid temperatures 
can be accurately determined by careful 
X-ray determinations on the solid solution 
limit. 



156 



PRECISION X-RAY STUDY OF ALUMINUM-SILVER ALLOYS 



The portion A B of the solubility limit 
represents an interesting application of the 
X-ray method. As is well known, the beta 
phase cannot be retained, while the alpha 
can be retained by quenching. Conse 
quently, films of two-phase alloys quenched 
from this temperature range show sharp 
lines from the alpha phase and blurred lines 
from the products of the beta decomposi 
tion. The X-ray method can be used there 
fore to determine accurately solubility 
limits of a quenchable phase, even though 
the phase with which it is in equilibrium 
cannot be retained by quenching. (Cf . Jette 
and Foote 22 on Fe-Ni, and Mathewson 23 on 
Ag-Hg alloys quenched from the alpha 
plus liquid regions.) In this temperature 
range, the solubility of aluminum in 
silver increases somewhat with decreasing 
temperature. 

The portion BC of the solubility curve 
represents equilibrium between the alpha 
and gamma phases, both of which can be 
retained by quenching. In this temperature 
range, the solubility of aluminum in 
silver decreases slightly with decreasing 
temperature. 

Below the peritectoid temperature (i.e., 
with beta prime as the adjoining phase), the 
solubility of aluminum in silver decreases 
rapidly with decreasing temperature. The 
filings for the X-ray analysis were annealed 
for long periods of time in order to approach 
true equilibrium as closely as possible. In 
all cases, equilibrium was approached from 
the same side; i.e., by the precipitation of 
beta prime from the alpha phase. The fact 
that widely different annealing times gave 
solubility limits that yield a smooth curve 
when plotted against annealing tempera 
ture indicates that the heat-treatments 
were adequate. Another criterion of equi 
librium is the sharpness of the X-ray lines. 
If the phases present are not of uniform 
composition, X-rays will be diffracted over 
a range of angles and the lines would be 
blurred. In all cases, sharp lines were 
obtained. 



Hoar and Rowntree 17 and Hume-Roth- 
ery, Mabbott and Evans 18 have determined 
the solid solution limit by microscopic 
methods. Of a total of 29 microscopic obser 
vations by these investigators only 5 fall 
in the wrong phase fields, and even these 
are very close to the solid solution limit 
as determined by X-ray methods. Thus, if 
the X-ray and microscopic methods are 
carefully and properly applied, the results 
on solid solubility determinations are in 
complete agreement. In other words, the 
phase diagram is truly a function of 
the system itself and is independent of the 
method used to investigate the system. 
Cases of such complete agreement between 
these two methods of investigation are not 
common. 

Ageew and Shoyket 11 have reported con 
siderably higher solubility limits as a result 
of X-ray and microscopic studies. Their 
alloys were given short anneals and it is 
very doubtful whether equilibrium had 
been attained. 

SUMMARY 

Precision lattice-constant measurements 
have been made on silver-rich alpha solid 
solution alloys of the silver-aluminum sys- 

TABLE 4. Summary of Solubility Limit 



Temperature, 
Deg. C. 


Atomic 
Per Cent 
Al 


Weight 
Per Cent 
Al 


Adjoining 
Phase 


779 


17.84 


4-67 




750 


18.18 


4-77 


Beta 


700 


18.85 


4.98 


Beta 


650 


19.62 


5-22 


Beta 


609 & 


20.34 


5-44 




550 


20.19 


5-40 


Gamma 


500 


20.06 


5-35 


Gamma 


452 


19.92 


5-31 




400 


17.02 


4.42 


Beta prime 


300 


12.52 


3-13 


Beta prime 


2OO 


8.75 


2.12 


Beta prime 



a Peritectic temperature, a + L <= 
ft Eutectoid temperature, ft + a + y 
c Peritectoid temperature, a + 7 *=* & 



tern. In the homogeneous field, the lattice 
constants are a linear function of the com 
position (in atomic per cent). Densities 



FRANK FOOTE AND ERIC R. JETTE 



157 



have been calculated from these lattice 
constants on the assumption of simple sub 
stitution of aluminum for silver atoms in 
the silver lattice. These calculated densities 
check the directly measured densities 
within the experimental error of the density 
measurements. The solid solubility limit 
has been determined over the temperature 
range of 200 to 779C. The limit thus 
established is in close agreement with micro 
scopic results reported by other investi 
gators. The results of determinations of the 
solid solubility limit are summarized in 
Table 4- 

REFERENCES 

i. M. Hansen: Aufbau der Zweistofflegierungen, 

1-5. Berlin, 1936. 1. Springer. 
2 I Obinata and M. Hagiya: Kinzoku no Kenkyu 

(1935) 12, 4IP-429; Chem. Abs. (193,6) 30, 66. 

3. W. Hofmann and K. E. Volk: Metallwirtschaft 

(1936) iSi 699-701. 

4. H. Taziri: Tetsu-to-Hagane (1938) 24, 357-370; 

Chem. Abs. (i939) 33, 4922. 



5. F. E. Tischtchenko: Jnl. Genl. Chem. (U.S.S.R.) 

(1939) 9i 729-731- 

6. E. R. Jette and F. Foote: Jnl Chem. Phys. (1935) 

3, 605-616. 

7. M. Siegbaum: Spektroskopie der Rontgen- 

strahlen, Ed. 2. Berlin, 1931. J. Springer. 

8. M. U. Cohen: Rev. Sci. Instr. (1935) 6, 68-74. 

9. W. F. Hillebrand and G. E. F. Lundell: Applied 

Inorganic Analysis, 389-401. New York, 1929- 
John Wiley and Sons, 

10. A. F. Westgren and A. J. Bradley: Phil. Mag. 

(1928) [7] 6,280-2?^. 

11. N. Ageew and D. Snoyket: Jnl. Inst. Metals 

(1933) 52 119-129. 

12. C. S. Barrett: Metals and Alloys (1933) 4> 63-64, 

74. 

13. R. T. Phelps and W. P. Davey: Trans. A.I.M.E. 

(1932) 99, 234-245. 

14. A. Phillips and R. M. Brick: Trans. A.I.M.E. 

(1934) in, 94-118. 

15. S. Kokubo: Sci. Repts. Tohoku Univ. (1934) 23, 

45-5I- 

16. E. R. Jette and F. Foote: Phys. Rev. (July 1940). 

17. T. P. Hoar and R. K. Rowntree: Jnl. Inst. Metals 

(1931) 45, 119-125. 

18. W. Hume-Rothery, G. W. Mabbott and K. M. C. 

Evans: Trans. Roy. Soc. London (1934) 233-A, 
66-70. 

19. G. I. Petrenko: Zlsch. anorg. Chem. (1905) 46, 

4059. 

20. E. Crepaz: Atti III Cong. naz. chim. pura appl. 

Firenze e Toscana (1929) 3, 371-379- 

21. F. E. Tischtchenko: Jnl. Genl. Chem. (U.S.S.R.) 

22. E. R. Jette and F. Foote: Trans. A.I.M.E. (1936) 

120, 259-272. 

23. H. M. Day and C. H. Mathewson: Trans. 

A.I.M.E. (1938) 128, 261-280. 



X-ray Analysis of Hot-galvanized Heat-treated Coatings 

BY F. R. MORRAL,* MEMBER A.I.M.E., AND E. P. MILLER t 

(Cleveland Meeting, October 1940) 



HOT-DIPPED heat-treated zinc coatings 
on sheet steel were examined with X-rays. 
The phases identified were compared with 
those listed in the modern literature of the 
zinc-iron system. The zeta (FeZnis) phase 
is shown to correspond to the phase found 
by Schueler (1925) in heat-treated zinc 
coatings. 

An investigation was undertaken to 
determine the phases present in hot-dipped 



Temperafure.deg.C. _ 




\ 




<\ 


\ 


1832 

1472 ^ 
o 

ttQ* 

1112 3 

If 1 

752 | 
392 
32 


A 


r\ 


^ 




\ 


"-x \ 


\/ 


8?3"C. 




f- 


6&TI 




. 


/ 










L J (, 


$ 
B 












r 


d 


* 




d 






e 


f 


ff / 


K 


e 20 40 60 80 Z> 



Zinc, per cent 

FIG. i. SYSTEM IRON-ZINC ACCORDING TO 
J. SCHRAMM. n 

heat-treated zinc coatings on steel sheets 
and wire. This kind of coating on wire was 
studied some years ago by Schueler 1 by 
microscopic and chemical methods. He 
reported a new phase, FeZnio. Later, Clark 
and Stillwell, 2 using X-ray crystal analysis, 
identified a new phase in ordinary galva 
nized wire, which was assumed to be the 
FeZnio phase. 

Manuscript received at the office of the Institute 
Sept. 13, 1939; revised May i, 1940. Issued as T.P. 
1224 in METALS TECHNOLOGY, September 1940. 

* Research Metallurgist, Continental Steel Corpo 
ration, Kokomo, Indiana. 

t Physics Department, Purdue University, Lafay 
ette, Indiana. 

1 References are at the end of the paper. 



PREVIOUS WORK 

Since the extensive review of the iron- 
zinc alloys presented a few years ago, 3 much 
new work has been published, some of 
which 5 " 16 was too late to be included in the 
recent edition (1939) of the American 
Society for Metals Handbook. 4 

Figs, i and 2 are reproductions of the 
latest iron-zinc diagram proposed by 



Temperature, deg. C . 

(Ji <T> 
O Q r 

5 o 


C 668" 


H r*S 


-V 


u." 


gj^ 


^ 


^.:- 

t \ * ~ 


ga?\p 




O { 


1 \ 

4 \ 


R 


l+s \ 

S30 S\ 


" L -g 

I 
f 

932 E 


| 


\ 


f- Tt me- temperature carves 
D Tempera tare -temperature 
difference curves 
Suscephbilify- lemperoiivre 
curves 


A 

w 


S+s 

4I< U 


& 

752 
n 


/ 


5^/7 vw 


5 10 5 Z 



158 



I ron ( percent 

FIG. 2. TRANSFORMATION OF 5 IN 5i IN IRON 

AND ZINC SYSTEM. 

Schramm 11 as a result of his investigations 
by microscopic, thermal, magnetic suscepti 
bility and X-ray methods. Table i sum 
marizes all the known published X-ray data 
of the phases present. 

The solubility of iron in zinc has been 
determined by X-ray methods. 10 In the 
same manner the homogeneity range of the 
F phase was studied 10 using the principle 
that the parameter of an intermetallic 
compound changes while the concentration 
of one of the constituents changes. 

Schramm 10 has found some So alpha lines 
in X-ray powder photograms of phase 5i, 
and from these he believes that this phase is 
related to the F phase. Previous investi- 



F. R. MORRAL AND E. P. MILLER 



159 



gators had considered this 81 phase to be 
hexagonal close packed 17 with an axial ratio 
/# = i.i3o. 12 It is not a superstructure, 
but a gigantic cell containing 550 + 8 
atoms in the unit cell. This would make it 
the largest cell found to date among the 
intermetallic phases. More work seems to 
be desirable on the structure of the 81 
phase. 24 The formula given is also uncer- 

TABLE i. Phases in 



although having lower symmetry. Halla, 
Weil, and GoetzePs 16 data for this phase 
are given in Table i. They established 
definitely that FeZnis is monoclinic. 
There are two molecules of FeZn^ per unit 
ceU. 

It has been mentioned before that a 
phase designated as FeZnio was determined 
some time ago 1 by chemical analysis and 

the Zinc-iron System 



Greek 

Letter 


Phases 


Formula, Per Cent 
Fe 


Crystal 
Structure 


Space 
Group 


Atoms 
Ele 
mentary 
Cell 


Parameter, A. 


Remarks 


X-ray 
Data 


At 
omic 


Wt. 


Alpha.. 


a Fe 


FeZn 






b.c.c. 


Oh 9 


2 


2. 862-2 .943 


Up to 40oC. 
dissolves 6 per 
cent Zn 


Gamma 


T Fe 


FeZn 






f.c.c. 


A5 


4 


Not deter 
mined 


May dissolve as 
much as 43 per 
cent Zn 


Capital 
Gam 
ma. 


r 


FesZn2i a 


23 2 

31 3 


20 5 

28 


b.c.c. 


T d 3 


52 


8.9560-8.9997 


10 




Si 


FeZn? 


8.1 
13-2 


7 
II. 5 


Hexagonal 


Cec 4 = Cemc 
or 

Deft 4 = Cemmc 


550 8 


a 12. 80 
c 57-6 
c/a i . 1300 

12 


Present between 
room temp, and 
640C. 

10 


Delta.. 


6 


FeZn? 


8.1 
II. 5 


7- 
10 








Not deter 
mined 


620-668C. 


Zeta.. . 


f 


FeZn is 


7-2 

7-4 


6.0 

6.2 


Monoclinic 


C. h * - C 2/m 


28 


a 13 65 
b - 7.61 
/9 I2844 
c 5.06 


There are two 
molecules to 
the unit cell. 

16 


Eta.... 


i? 


Zn 




Max. 
0.003 


Hexagonal 
c.p. 


DBA 4 


2 


a 2 . 6600 
& 4-9379 
c/a i , 8563 


4u 



a The T phase occasionally has been given the formulas FeZns, FeaZnio 3 - 16 and FesZnai. 18 - 19 This last 
formula is to be preferred although it is not consistent with any regular distribution of the two kinds of metal 
ions between the points of the basic alpha-brass structure, 20 whose characteristic is the ratio of 21 electrons to 
13 atoms. W. L. Bragg 21 has stated that the relation between number of electrons and atoms in^the structural 
unit overrides the influence of regular arrangement. Jones, 22 using the reasoning of wave mechanics, has made a 
study of the energies of the loosely bound electrons in the alloys with the structure and found that the ratio of 
211 13 comes very near to completely filling a zone. The magnetic properties are found to agree with such an 
arrangement. Thus, although the actual amounts of iron_and zinc may not correspond to the relationship 
FeaZnai, it has the characteristic arrangement of atomic sites on which the conception of a phase is formed. 
In other words, Hume Rothery s rule 20 and Jones theory,22 although indicating a definite number of atoms in 
the unit pattern, make no demand that the ratio of the atoms of different kinds should be represented by simple 
integers. 23 



tain. No X-ray work has been done on the 
structure of the high-temperature delta 
phase. 

Zeta, 8 a new phase with a narrow range 
of homogeneity (6.0 to 6.2 per cent Fe by 
weight) has been established by thermal, 
microscopic, magnetic susceptibility and 
X-ray methods. The crystal structure was 
claimed 10 to resemble that of the T phase, 



found in large amounts in heat-treated zinc 
coatings. This phase was believed to corre 
spond to the C phase found in galvanized 
wire coatings by X-ray diffraction. 2 A com 
parison of the Clark and Stillwell data for 
C phase (7 lines) with that of Schramm 8 
(52 lines) suggests the identity of this phase 
with FesZn2i. The first two investigators 
maintain that the new phase is body- 



l6o X-RAY ANALYSIS OF HOT- GALVANIZED HEAT-TREATED COATINGS 



centered cubic, and the German author 
believes that it is cubic of low symmetry. 

EXPERIMENTAL PROCEDURE 

Samples from a commercial galvanized 
sheet taken at random were submitted to 
X-ray diffraction analysis. The sheet was 
hot-dipped in a "double-bottom" galva 
nizing pot 25 - 26 and the coating was a gal- 



solution of sulphuric acid to remove any 
superficial zinc oxide. After this treatment 
the weight of coating was 0.58 oz. per 
square foot of sheet. The analyses of the 
steel and of the coating are given in Table 2. 

X-RAY DIFFRACTION ANALYSIS 

The X-ray diffraction pictures were taken 
by allowing collimated unfiltered radiation 



Fe- 



Zn-rj 



l i 



h 



Hof- galvanized 
(Galvotnneal) 



$(FeZn 13 ) 



\ h 



i. 







10 



60 



70 



FIG. 3. 



20 30 40 50 

Angle of reflection, 6, deg. 
COMPARISON OF COMPOSITE X-RAY DIFFRACTION PATTERN OF HEAT-TREATED HOT- 

GALVANIZED COATING AND PATTERNS OF KNOWN IRON-ZINC ALLOYS. 



vannealed," 1 - 27 a commercial process that 
consists in passing the freshly galvanized 

TABLE 2. Analyses of Steel and of Coating 





"Gal- 




"Gal- 


Sample 


van- 
nealed" 


Sample 


van- 
nealed" 




Sheet^ 




Sheet 


I. Analysis of base 




II. Weight of 




steel, per 




coating, oz. 




cent: 




persq. ft 


0.62 


Carbon ...... 


o 02 


III. Analysis of 




Manganese. . . . 


0.40 


coating, per 




Phosphorus 


0.085 


cent: 




Sulphur 


o 057 


Iron. 


10 O2 




Tr 


Lead 


i . 07 


Copper 


o 30 


Tin 


o. 34 


Nickel 


o. 26 


Zinc (by dif 








ference) 


88.57 



sheet through a muffle furnace at i2ooF. 
for approximately 15 sec. when traveling 
at 60 ft, per minute. The sample was dipped 
for one minute in a 1,7 per cent by volume 



to fall on the sheet sample at a glancing 
angle with its surface. The diffraction pat 
tern was obtained on a film located on a 
cylindrical camera placed about the sample. 
The diameter of the camera used in this 
work was 15.229 centimeters. 

The sample was examined at different 
glancing angles, to bring different regions 
of the picture into focus at different times. 
From these a composite table was made 
(Table 3). This composite picture was 
verified by making an X-ray diffraction 
analysis on heat-treated galvanized wire.* 1 
The intensities of the lines were hard to 
evaluate on a composite diffraction photo- 
gram. They are indicated in Table 3 with 



* The wire readings had to be corrected to take into 
consideration the fact that the diffraction occurs from 
several points on the surface. 2 



P. R. MOE.RAL AND E. P. MILLER 



161 



reservations. The angles indicated in the 
table refer to cobalt-K ai radiation to permit 
comparison with published work. 10 

TABLE 3. Diffraction Pattern of Heat- 
treated Zinc Coatings 
ANGLES CALCULATED TO Co-K a RADIATION 



Radiation 
Intensity 


Radia 
tion 
Angle 


dhki, 
A. 


Source 


Strong . . 


18 20 
20 36 

23 17 


2.84 
2-53 
2. 197 


Zeta (FeZma) 
Zeta and FesZmi 
Zeta and FesZn2i 


Medium 


24 23 


2. l6l 


Zeta 


Strong 


24 43 


2. 134 


FeZn? and FesZnai 


Weak . 


25 


2. 112 


Zeta and FesZnai 


Medium 
Very weak. . . 
Weak 


25 17 
25 33 
26 9 


2.09 

2.07 

2.025 


Zinc (2.08) 
Zeta 
Zeta 


Weak 


26 40 


1.988 


Iron (no) 2.01 


Very, very 
weak . . 


27 42 


I 922 


Zeta 


Very weak. . . 
Very, very 
weak 


27 46 
28 18 


I-9I5 
1.880 


Fe 6 Zn 2 i 
FesZmi 


Very, very 
weak 


28 34 


1.866 


Unidentified 


Very, very 
weak 


28 54 


i. 847 


Zeta 


Very weak. . . 
Weak 


29 24 
30 44 


i 818 

i . 746 


Zeta 
Zeta and FesZnai 


Weak . ... 


32 3 


1.682 


Zeta 


Weak 
Weak 
Weak 


32 29 
32 52 
34 26 


1.661 

1.644 
1.578 


FeZn? (1.67) 
FeZn? (?) (1.626) 
Zeta 


Very, very 
weak 
Weak 


36 19 
36 46 


1.506 
1.491 


FeZn 7 (1.518) 
FeeZnai 


Strong 


38 34 


1.431 


Fe (200) and FeeZmi 


Weak 


39 14 


1.411 


Zeta 


Weak 


40 10 


1-383 


Zeta and FeZnr (1.393) 


Very weak.. . 
Weak 


41 9 
42 33 


1.356 
1.320 


Zeta 
Zeta 


Weak 


42 57 


1.310 


FesZnn 


Very, very 
weak 
Medium 
Weak 


43 21 
44 17 
44 SO 


1.300 
1.278 
1.266 


Fe 5 Zn 2 i 
Zeta and FeZn? (1.274) 
Zeta and FeZn? (1.25) 


Weak 


46 33 


i. 229 


FesZn2i (1.213) 


Very weak. . . 

Medium 
Very weak. . . 
Very; weak. . . 
Medium 


47 6 

SO 
56 58 
57 34 

62 2 


1.218 

1.165 
1.064 
1.057 

1. 010 


Zeta, FesZnai, and 
FeZn? (1.217) 
Fe (211) 1.166 
Zeta 
Zeta and FegZnsi 
Fe (220) i. 01 



In Fig. 3 a comparison is made of the 
data of Schramm 10 with the diffraction data 
from the present sample. Such a chart does 
not give an accurate check on the work. It 
permits only a rapid graphic survey of the 
various phases present. 

DISCUSSION OF RESULTS 

It is evident from the data in Table 3 and 
Fig. 3 that all of the intermediate phases 
of the iron-zinc system are found in the 



heat-treated zinc coating. The zeta phase is 
predominant. Schueler 1 found by chemical 
analysis that 60 per cent of the heat-treated 
galvanized coating consisted of a new phase, 
which he designated FeZnio. From the dif 
fraction data presented here and the above 
fact, it seems evident that the phase found 
by Schueler and the zeta phase (FeZnis) 
are the same intennetallic compound. 

The chemical analysis of the heat-treated 
zinc coating showed 10 per cent of iron, 
which is somewhat high for such a coating, 
the average analysis being about 8 per cent. 
The homogeneity range of the zeta phase 
is 6 to 6.2 per cent Fe. A relatively large 
amount of zeta material would be expected, 
therefore, the remaining iron being taken 
care of by FeZn? (7 to 10 per cent Fe) and 
FesZn 2 i (20.5 to 28 per cent Fe). 

The phases present in heat-treated zinc 
coatings have been checked on wire, al 
though the data presented here were ob 
tained from a sheet sample. 

SUMMARY 

The recent investigations on the iron- 
zinc system are summarized. 

The X-ray diffraction analysis of heat- 
treated zinc coatings on sheets (and on 
wire) shows that the alloy layers consist 
mainly of the zeta (FeZn 13 ) phase. Zinc, 
small amounts of FesZ^i and 5i phase are 
also present in the coating. 

The zeta phase corresponds to the phase 
found by Schueler in 1925. 

REFERENCES 

1. J. L. Schueler: Trans. Electrochem. Soc. (1925) 

47, 210. 

2. C. W. Stillwell and G. L. Clark: Ind. and Eng. 

Chem. t Anal. Ed. (1930) 2, 266. 

3. E. C. Truesdale, R. L. Wilcox and J. L. Rodda: 

Trans. A.I.M.E. (1936) 122, 192. 

4. E. A. Anderson, J. L. Rodda and G. Edmunds: 

Amer. Soc. Metals Handbook (1939) 1745. 
4a. E. A. Anderson and J. H. Craig: Amer. Soc. 
Metals Handbook (1939) 1757- 

5. J. Schramm, W. Heike, and O. Vaupel: Metall- 

wirtschaft (1936) 28, 655. 

6. H. Grubitsch and F. Bruckner: Korrosion and 

Metallschutz (1937) 13, 254. 

7. E. Scheil: Ztsch. Metallkunde (1937) 29, 209. 

8. J. Schramm: Ztsch. Metallkunde (1937) 29. 222. 

9. E. Scheil and H. Wurst: Ztsch. Metallkunde (1938) 

30, 4. 

10. J. Schramm: Ztsch. Metallkunde (1938) 30, 122. 
ir. J. Schramm: Ztsch. Metallkunde (1938) 30, 131, 



X-RAY ANALYSIS OF HOT-GALVANIZED HEAT-TREATED COATINGS 



12. H. Bablik, F. Goetzel and F, Halla: Ztsch. 

Metallkunde (1938) 30, 249. 

13. J. Schramm: Ztsch. Metallkunde (1938) 30, 333- 

14. H. Bablik: Korrosion und Metallschutz (1938) 14, 

168. 

15. H. Grubitsch and F. Bruckner: Korrosion und 

16. F. Halla, R. Weil and F. Goetzel: Ztsch. Metall 

kunde (1939) 31, 112. 

17. A. Osawa and Y. Ogava: Ztsch. Kristallographie 

(1928) 68, 177- 

1 8. A. Westgren and W. Ekman: Arkiv Kemi, 

Mineralogi och Geologi (1930) lo-B (n). 

19. W. Ekman: Ztsch. Phys. Chem. (1931) I2-B, 57- 

20. W. Hume-Rothery: Structure of Metals and 

Alloys, Inst. Metals Mon., London, 1936. 

21. W. L. Bragg: The Crystalline State, i, 153. 

London, 1933. 

22. H. Jones: Proc. Royal Soc. (1934) I44-A, 225. 

23. W. L. Bragg: Proc. Inst. Metals, May lecture, 

1935, 298. 

24. Discussion: Jnl. Amer. Zinc Inst. (1938) 19, 100. 

25. R. J. Wean: JnL Amer. Zinc Inst. (1938) 19, 94- 

26. R. J. Wean: Steel (1938) No. 7, 57- 

27. J. L. Schueler: Metal Progress (1037) 31, 501. 

DISCUSSION 

(P. H. Brace presiding) 
P. R. KOSTING,* Watertown, Mass. It is 
by no means a simple procedure to use X-ray 
for the identification of compounds in a com 
plicated mixture. The work of Messrs. Morral 
and Miller is therefore very interesting and 
constructive. In the identification of any com 
pound, it is not only necessary to match d 
values in any pattern, but also to check inten 
sity values. For instance, for iron we anticipate 
d values around 2.0, 1.4, 1.2, and i.o, and / 
values of strong, weak, medium, and weak, 
respectively. We note on Table 3 that the 
intensity of the 2.0 line for iron is weak, but 
the intensities of the 1.2 and i.i lines are 
medium. Similarly for zinc; normally we 
expect d values of 2,5, 2.3 and 2.1 with corre 
sponding / values of weak, weak, and strong, 
whereas on Table 3 there is only a medium line 
corresponding to the 2.1 d value. Traces of the 
2.5 and 2,3 line would be expected if the 2.1 line 
is of medium intensity unless grain size and 
orientation, etc., interfere. Such discrepancies 
in the anticipated intensity of lines should be 
pointed out and satisfactorily explained. 

W T . D. FORGENG,! Niagara Falls, N. Y. The 
authors have mentioned the lack of correlation 
between the results of Osawa and Ogava, Still- 
well and Clark, and Schramm and other recent 
German investigators, and at the risk of 
adding further to the confusion that exists in 
the X-ray data on the iron-zinc system, and 
particularly that concerned with galvanized 

* Rest-arch Metallurgist, Watertown Arsenal. 

f Union Carbide and Carbon Research Labs., Inc. 



coatings, I am presenting diffraction patterns 
and photomicrographs taken at different 
depths in a dip-galvanized coating. The diffrac 
tion patterns in Fig. 5 were taken with filtered 
cobalt radiation, using the metallographic 
specimen of Fig. 4. The specimen was polished 
to increasing depths, and the etched structure 
recorded. It was then transferred to the X-ray 
camera, tilted 10 into the X-ray beam, and the 
diffraction pattern was registered. 

The structures shown in the photomicro 
graphs are not necessarily represented by the 
diffraction patterns, since the X-rays penetrate 
the specimen to some extent. Some 35 to 60 
lines (some too faint to register in a print) are 
present in each pattern between the limiting 
6 values of 15 and 65. While some of the lines 
could be accounted for on comparison with 
Schramm s charts in a manner similar to that 
used by the authors, the agreement was far 
from satisfactory. Furthermore, an attempt to 
arrange the lines into a series of patterns by 
statistical methods required an assumption of 
at least n individual species, giving rise to 
separate diffraction patterns, whereas only five 
phases have been recognized microscopically. 
It is believed that these results can best be 
explained by the formation of solid solutions 
between the recognized constituents in the 
system, which result in a series of related but 
distinctly different patterns. This would also 
explain the lack of agreement in the published 
data referred to in the paper. 

F. R. MORRAL (author s reply). Cn page 
1 60 the difficulty of evaluating the intensities 
of lines, particularly on a composite diffraction 
photogram, is pointed out. We fully realize 
Mr. Kosting s point; namely, the importance 
that not only the d values match but also their 
intensities. 

Only the strongest line for zinc (d = 2.1; 
/ = strong) showed in the range of our photo- 
gram as medium (recorded in Table 3 as 
d = 2.09; / medium). The other zinc lines that 
are weak were outside the glancing angle for 
which the intensities could be estimated with 
any degree of satisfaction. 

Little work ( 2) has been published on X-ray 
diffraction analysis of galvanized coatings. It is 
unfortunate that Mr. Forgeng has not pre 
sented the data in more detail with his photo 
micrographs and the diffraction photograms. 



DISCUSSION 



163 




FIG. 4. DIP-GALVANIZED ; COATING POLISHED TO DIFFERENT DEPTHS PARALLEL TO STEEL 

BASE. X 250. 

a, outside surface (zinc -f- zeta?). 

6, adjacent to a (zeta + delta + zinc?). 

c, adjacent to b (zeta -f delta + capital gamma?). 

d, layer immediately adjacent to steel base (capital gamma?). 

2 per cent Nital etch. 



1 64 X-RAY ANALYSIS OE HOT- GALVANIZED HEAT-TREATED COATINGS 




f 



f 



> 5. DIFFRACTION PATTERNS TAKEN AT GLANCING ANGLE ON SPECIMEN OF FlG. 4- COBALT 

RADIATION. 

b, adjacent to a. c, adjacent to 6. d, adjacent to steel base. 



a, outside surface. 

TABLE 4. Comparison of Galvanized 
Coatings 





Galvanized 


Hot Dip 


Hot Dip, 

Heat- 
treated 


Weight of coating (approx.), oz. 


1.25 
o.ooi 

18 
30 

2 

SO 


0.60 
0.0005 

3 

27 
60 
10 




Phases by microscopic determi 
nation, 9 per cent: 

PeZn 




re ru, 





J. L. Bray and F. R. Morral: Zinc Coatings. Amer. 
Soc. Metals Preprint 45 (1940)- Symposium on Sur 
face Treatment. 



A regular hot-dip galvanized coating differs 
from the coating discussed in the paper in 
several respects (Table 4) . 

Mr. Forgeng s photomicrographs show well 
that these alloys are not evenly distributed 
in one layer but actually overlap. This effect, 
together with the penetration of the X-ray 
beam, causes the diffraction of more than one 
phase on the photographic plate. When two 
or more phases are formed on a photographic 
plate, the intensities present depend on the 
relative amounts of the phases present. This 
may account for the difficulty in making his 
data agree with those of Schramm. In addition, 
when two or more phases are present together, 
one of them may be slightly displaced. 



Creep and Recrystallization of Lead 



BY ALBERT A. SMITH, JR.,* MEMBER A.I.M.E. 
(Cleveland Meeting, October 1940) 



THE creep properties of metals have as 
sumed increasing importance in recent years 
and many investigations have been made 
on various phases of the problem. In the 
past year the annual lectures of the Insti 
tute of Metals Division 1 and the Iron and 
Steel Division 2 were concerned with this 
subject, and are excellent reviews. For many 
years the creep properties of lead and lead 
alloys have been investigated at the Central 
Research Laboratories of the American 
Smelting and Refining Co., and it is hoped 
that the present paper will contribute 
knowledge of some of the factors that influ 
ence creep. Lead, being a metal that recrys- 
tallizes readily at room temperature, is an 
excellent material upon which to investi 
gate creep phenomena because of the sim 
plicity of the apparatus needed, and it 
might be expected that results would apply 
to some extent to elevated-temperature 
tests of other metals. 

MATERIAL 

The creep tests of this investigation were 
all made on high-purity lead (99.9998+ 
Pb) in which the only impurity that could 
be detected by spectrographic and chemical 
means was approximately o.oooi per cent 
Fe. Flat strips, 0.750 in. wide by o.iooin. 
thick, were prepared by extruding 2 -in. dia. 
cast billets at a temperature of 25oC. This 
gave a material of 2 to 3 mm. average grain, 
size with a tensile strength of 1400 Ib. per 
sq. in. when tested at a straining rate of 5 

Manuscript received at the office of the Institute 
Jan. 31, 1940. Issued as T.P. 1227 in METALS TECH 
NOLOGY, September 1940. t 

* Research Metallurgist, American Smelting & 
Refining Co., Barber, N. J. 

1 References are at the end of the paper. 



per cent per minute. With this extremely 
soft material, great care was necessary in 
handling the specimens, to avoid distortion 
and subsequent recrystallization. 

APPARATUS 

The creep tests at 3oC. were conducted 
in a constant-temperature room, specially 
built for the purpose in the basement of an 
office building in which there was no mov 
ing machinery, and no vibration was de 
tectable under the most severe conditions 
of shock in the adjacent surroundings. The 
room was well insulated and heated elec 
trically, control being by means of a de 
Khotinsky regulator operating a relay. 
Maximum temperature variation over a 
period of several years has been KC. 

Specimens of the extruded strip, 16 in. 
long, were used for test purposes with no 
reduced gauge section, thus eliminating any 
distortion due to machining. The method 
of clamping and supporting the specimens 
is shown diagrammatically in Fig. i. 

The gauge marker for the measurement 
of extension consisted of an aluminum 
strip, 0.050 in. thick by 0.75 in. wide, one 
end of which was ground and sanded to a 
precise knife-edge and fastened to the speci 
men by suitable pins. After the strip was 
fastened in place a fine scratch was in 
scribed just below the knife-edge by means 
of a sharp teasing needle. Just below this 
gauge scratch a short vertical mark was 
made so that the same point could be 
located and read each time. Several hun 
dred specimens have been prepared by this 
method, and it was found that with experi- 
165 



i66 



CREEP AND RECRYSTALLIZATION OF LEAD 



ence excellent gauge marks could be made. 
As far as could be ascertained by careful 
observation, no shifting of the aluminum 
strips has taken place over test periods up 
to three years. 



Support 



Lead 
specimen- 



>"C"C/amp 
k-flott steel grips 

^ Aluminum 
gauge marker 



^"Knife-edge 
"C Clamp 



$&^ flat steel grips 
^Centering pin 

^Spring 



FIG. i . SPECIMEN ASSEMBLY. 

Strain measurements were made by de 
termining the distance between the lower 
edge of the aluminum strip and the fine 
scratch inscribed on the lead specimen by 
means of a portable micrometer microscope 
sensitive to 0.000025 in. Actually the limit 
of sensitivity depended upon the fineness of 
the cross scratch on the lead specimen and 
it is estimated that the accuracy was at 
least 0.0005 i n - Since the gauge length 
was 10 in., this gave a minimum accuracy 
of 0.00005 * n - P er i nc h. 

The tests at 55C. were conducted in an 
electrically heated, thermostatically con 
trolled oil bath, which contained 12 test 
stations. In a tank 4 in. wide, 86 in. long 
and 22 in. high, with a glass front, clear 
transformer oil was used, through w T hich 
readings could be made with the micro 
scope. The specimens were connected at the 
bottom to an angle beam welded to the 
tank. Loading was accomplished by tak 



ing the load out at the top of the tank by 
means of a lever arm with a one to one 
ratio. Readings were secured by the use 
of aluminum markers in the same manner 
as described above for the 3oC. tests. 




100 200 f 300 400 500 600 

Time .days 
FIG. 2. CREEP AT 30C. 

Exceptionally good temperature control 
was maintained in the tank, variation being 
only about %C. at the maximum. 

A similar tank was used for the iooC. 
tests but air heating was employed because 
the oil discolored at this temperature to 
such an extent that readings could not be 
made. Air, preheated externally by resist 
ance heaters, was circulated at high veloc 
ity through the tank to secure as uniform 
heating as possible. The temperature vari 
ations with this tank w r ere about 4C. 

DATA 

Time-extension curves for the 30 C. tests 
are plotted in Fig. 2, and it is apparent 
that in some cases creep does not proceed at 
a uniform rate but rather by intermittent 
steps. Curve G is particularly interesting on 
this point. For the first 15 days of the test 
there was no discernible motion (less than 
0.005 P er cent) but the next reading, at 18 
days, showed 0.025 P er cent extension. The 
extension proceeded somewhat irregularly 



ALBERT A. SMITH, JR. 



I6 7 



for 231 days and then for a period of 140 
days there was no further motion. At 371 
days the specimen again started to elongate, 
the rate accelerating for a time and then 




50 100 150 200 250 300 

Time, days 
FIG. 3. CREEP AT 55C. 



decreasing. Obviously, it is difficult to de 
termine creep rates from these stepped 



sA - 249 Ib.persq. in. 
/, B-220ib.persqJn. 
JC-199.5 Ib.Dersa.in. 




25 50 75 100 125 150 

Time, days 
FIG. 4. CREEP AT iooC. 

curves, but a smooth curve can be drawn 
through the high points as shown by the 
solid line of curve G. 



Figs. 3 and 4 show time-extension curves 
for 55 and iooC. of other specimens at 
low stresses, which are also of the step type 
and on which smooth curves can be drawn 



500, 
E 400 

5T300 
250 
J"200 

150 
^100 



ES&_ 



0.01 



10,000 



0.10 1.0 10 100 1000 

Per cent per year, log scale 
FIG, 5. LOG-RATE LOG-STRESS CURVES. 



through the high points as shown by the 
solid lines. Chronologically there was no 
relationship between the various specimens 
as to the time when rapid motion would 
take place. It is evident that a sudden shock 
should affect in the same manner two ad 
jacent specimens suspended from the same 
angle support, nor can this explain long 
periods of zero extension. From this evi 
dence and from examination of creep data 
from hundreds of other tests, it is believed 
that creep may occur intermittently and 
that the periods of rapid and slow extension 
take place at random. Also, it is very prob 
able that any vibration superimposed on 
the tensile stress may produce a more uni 
form type of elongation. 

If the minimum creep rates obtainable 
from the time-extension curves are plotted 
on a log- rate log-stress scale as in Fig. 5, 
continuous lines are obtained for each of the 
three temperatures. At the higher stresses, 
there are changes of slope of the curves, 
indicating a higher rate of extension than 
might be expected by extrapolation from 
the low stress region. Microscopic examina 
tion of the specimens corresponding to the 
high-stress part of the curve was made 
immediately before and after the creep 
tests, and it was found that partial or com- 



i68 



CREEP AND RECRYSTAIXIZATION OF LEAD 



plete recrystallization had taken place. The 
resulting grain size, after deformation of 
about 3 per cent, was 5 to 20 times as large 
as the original grain size, so that there was 



To confirm this point, a new specimen 
(curve D of Fig. 2) was carefully etched and 
then a Crosshatch of very fine lines spaced 
about J-s in. apart was inscribed on one flat 




FIG, 6. CREEP SPECIMEN. 
a, original etch ; b, same area re-etched. 



X50- 



little question that the specimens had re- 
crystallized during the test. On the other 
hand, specimens C and E at 3oC., D and E 
at 55C. ? and C 3 D and E at iooC., corre 
sponding to the low-stress region, showed 
no change in grain size and it is doubtful 
whether any recrystallization had taken 
place. 



surface by means of a razor blade. The 
specimen was then loaded at a stress of 260 
Ib. per sq. in. and creep readings were made. 
In 171 days the specimen extended 3 per 
cent and this particular curve shows a 
period of about 10 days in which there was 
practically no motion and then a rapid 
extension of 0.35 per cent in 7 days, after 



ALBERT A. SMITH, JR. 



169 



which the extension proceeded as though no 
halt had occurred. 

After the test -was completed various 
parts of the surface of the specimen were 



original grain structure was still present 
with no evidence of recrystallization. 

In order to determine the time necessary 
to produce recrystallization at 3oC, by 




FIG. 7. TENSILE SPECIMEN. 
a, original etch; b, same area re-etched. X 5- 



photographed with the original etch, and 
one of these photographs is shown in Fig. 
6a. Four weeks after the completion of the 
creep test the surface was heavily re-etched, 
the region previously photographed was 
relocated by counting the cross lines and 
photographed again. Fig. 6b shows that the 



straining in tension, samples from the strip 
adjacent to those used for creep tests were 
extended i and 3 per cent in a testing 
machine at a rate of 5 per cent per minute. 
From microscopic examination, it was found 
that the sample extended 3 per cent com 
pletely recrystallized in about one hour. 



CREEP AND RECRYSTALLIZATION OF LEAD 



Fig. ya shows the sample immediately after 
straining and Fig. 76 shows the same area 
after recrystallization had taken place. It is 
evident that there is a decided difference in 
the grain size produced by the small de 
formation. The sample that was extended 




0.0026 0.0026 Oj0030 0.0032 0.0034 



FIG. 8. DATA ON PURE LEAD. 

i per cent recrystallized in about 10 days 
and the resulting grain size was somewhat 
larger. The photograph of the creep speci 
men, Fig. 6, was taken four weeks after the 
test had been completed and four months 
later there was still no visible change in the 
grain structure. 

It is therefore concluded that the low- 
stress section of the curves of Fig. 5 repre 
sents creep with very little or no strain of 
the type that produces recrystallization, 
whereas the higher creep rates, correspond 
ing to the high-stress section of the curves, 
are accompanied by recrystallization. 

The photomicrographs show many slip 
(or deformation) bands, although unlike 
normal slip lines even the extremely deep 
re-etching, which produced many pits, did 
not completely eliminate them. In some 
cases these bands became curved as they 
neared grain boundaries. This particular 
phenomenon has been noted by Hanson 
and Wheeler 3 on creep specimens of alumi 
num and they suggest that the bands are 
determined by maximum shear stresses 
rather than by crystallographic directions. 



DISCUSSION OF RESULTS 

The creep data presented show that 
creep does not necessarily take place in a 
smooth, continuous manner, but may be 
intermittent, with periods of rapid and 
slow extension. There does not appear to be 
any relationship between time or amount of 
extension relative to the occurrence of these 
steps; apparently they take place at ran 
dom. Recrystallization is not necessarily a 
factor, 4 as none could be detected except at 
comparatively high rates of extension, 
whereas the steps occur mainly at low 
rates. Considering the high purity of the 
lead used, any strain-aging phenomenon 
must be excluded. In a recent discussion 5 
based upon the ideas of Orowan 6 it was 
suggested that the occurrence of stepped 
creep curves is determined by the simul 
taneous effects of stress concentration and 
thermal agitation. At some certain time a 
locally concentrated stress is further in 
creased by the haphazard thermal stress 
fluctuations in the lattice and sudden elon 
gation results. It was postulated in the dis 
cussion at that time that steps in the creep 
curves might be larger if very pure mate 
rials and lower temperature and stresses are 
used. In general this has been found to be 
true and these data indicate that at iooC. 
the steps are less pronounced. Another 
point observed in the study of the creep of 
lead is that the steps also tend to become 
less pronounced with decreasing grain size. 

The change in the slope of the stress-rate 
curves of Fig. 5 is of considerable interest 
and has been noted by other investigators, 7 * 8 
who refer to it as the change from plastic to 
viscous flow. The present results indicate 
that a major difference between the two 
regions is the presence or absence of recrys 
tallization. At the high rates of extension, 
creep is accompanied by recrystallization. 
At the low rates the deformation appears to 
be confined in a large degree at or near 
the grain boundaries, although there is 
considerable internal deformation, as the 



DISCUSSION 



171 



photomicrographs indicate. However, the 
mechanisms of both the grain boundary and 
internal deformation are of such a nature 
that recrystallization does not occur. As the 
temperature decreases, the creep rate at 
which the transition from one type of flow 
to the other occurs also decreases. 

Kanter 9 has proposed an energy analysis 
of the creep process from which an energy 
value can be calculated that is of the same 
order of magnitude as the heat of diffusion 
of the metal. He plots the log flowability 

/ rate \ 

I 1 versus the reciprocal of the abso- 

\stress/ 

lute temperature, and from the resulting 
straight lines computes an energy value 
that is approximately equal to the heat of 
diffusion. 

The data on pure lead are plotted in this 
manner in Fig. 8, the upper three lines for 
ioo, 150 and 200 Ib. per sq. in. being from 
the low rate part of Fig. 5 and the other 
curve for 350 Ib. per sq. in. from the high 
rate portion. In the first instance an energy 
value of 16,500 cal. per gram mol is ob 
tained and for the other case a value of 
21,000 cal. per gram mol. The accepted 
value of the heat of self-diffusion of lead is 
28,000 cal, per gram mol, 10 which is con 
siderably different from the value com 
puted from the low-creep-rate data. The 
value from the high rate of extension is 
more nearly comparable but this is at a rate 
where spontaneous recrystallization takes 
place and the process cannot be closely 
related to diffusion. 

CONCLUSIONS 

1. The creep properties of high-purity 
lead have been determined at temperatures 
f 3> 55 an d iooC. 

2. It has been shown that creep of lead 
does not necessarily proceed uniformly but 
may take place in intermittent steps. 

3. At comparatively high rates of exten 
sion recrystallization occurs while at lower 
rates none could be detected. When recrys 
tallization is a factor, the creep rates are 



greater than would be obtained by extra 
polation from the low rate part of the log- 
rate log-stress curve. 

4. Employing the method proposed by 
Kanter, an energy value of 16,500 cal. per 
gram mol is obtained that is considerably 
different from the value for the heat of 
self-diffusion of lead. 

ACKNOWLEDGMENT 

The author is indebted to Dr. A. J. 
Phillips, under whose direction this work 
was done, and to Mr. P. A. Beck, for their 
helpful suggestions in preparing this paper. 

REFERENCES 

1. D. Hanson: Trans. A.I.M.E. (1939) 133, 15. 

2. H. W. Gillette: Trans. A.I.M.E. (1939) 135, 12. 

3. Hanson and Wheeler: Jnl. Inst. Metals (1931) 

45, 229. 

4. Greenwood and Worner: Jnl. Inst. Metals (1939) 

64, 135. 

5. Smith and Beck in discussion of paper by 

M, Gensamer: Trans. A.I.M.E. (1938) 131. 

6. E. Orowan: Ztsch. Physik (1935) 97 573. 

7. B. Chalmers: Proc. Royal Soc. (1936) is6-A, 427. 

8. Hanf stengel and Hanemann: Ztsch. Metallkunae 

(1938) 30, 41- 

9. J. Kanter: Trans. A.I.M.E. (1938) 131,385. 

10. Seith and Keil: Ztsch. Metallkunde (1933) 25, 104. 

DISCUSSION 

(B. B. Betty presiding) 

J. J. KANTER,* Chicago, III The data 
offered by Dr. Smith upon the creep character 
istics of pure lead are of considerable theoretical 
interest. It seems reasonable that at the small 
loads and slow rates of creep, the strain curve 
need not be smooth, in view of the coarse 
grained material used. It may well be that the 
strain accumulates intermittently through two 
types of movement, one associated with the 
grain boundaries; the other with slip planes. 

In connection with Fig. 8, it is of interest to 
consider the data of Hanffstangel and Hanne- 
man represented in the accompanying diagram 
(Fig. 9), where a constant strain rate at 70.4 Ib. 
per sq. in. is represented as a function of a 
reciprocal of the absolute temperature. An 
energy value of 13,500 cal. per gram mol is 
deduced from these data and compares favor 
ably with the 16,500 cal. per gram mol from 
Dr. Smith s data. Incidentally, it might be 
mentioned that had Dr. Smith plotted log 
rate of strain vs. reciprocal of absolute tempera- 

* Crane Company. 



172 



CREEP AND RECRYSTALUZATION OF LEAD 



ture, instead of log flowability, the energy value 
so deduced would be identical with that from 
Fig. 8. Since, for the data under consideration, 
stress enters into the calculation of each curve 



50 



Temperctf ure, cleg. F. 
100 150 200 



250 



Q = /3,500 Calories per gram mot 




3.5 3.0 2.5 

Reciprocal absolute temperature , sj< * JO 3 

FIG. 9. RECIPROCAL ABSOLUTE TEMPERA 
TURE PLOT OF HANTFSTENGEL AND HANNEMAN 
CREEP DATA FOR SOFT LEAD UNDER A STRESS 
OF 70.4 LB. PER SQUARE INCH. 

Drawn from data found in paper by K. von 
Hanffstengel and H. Hanneman: Mechanism of 
Creep and Fatigue Limit Investigated on Lead 
and Lead Alloys. Ztsch. Mctallkunde (1937) 
29(2), 50-52. 

as a constant term, the only advantage for the 
log-flowability scheme is if a constant stress 
curve is to be deduced from data involving a 
variety of stresses where it is necessary to find a 
stress function that will permit the data 
to be compared at picked constant stress 
values. 

The fact that the energy value deduced in 
the data falls short of the heat of self -diffusion 
for lead possibly is still reconcilable with the 
idea that some aspects of creep are akin to solid 
diffusion. It still may be contended that this 
energy value is of the order of magnitude of 
the heat of self -diffusion. When one takes into 
consideration the fact that the crystal move 
ment associated with this value is largely in 



the locale of grain boundaries, it would seem 
that the restraints upon a process of solid 
diffusion so taking place would be appreciably 
less than solid diffusion through an orderly 
space lattice. A grain-boundary atom might be 
considered to belong almost as much to the 
adjoining crystal as to the parent crystal, and 
thus the activation energy necessary to pro 
mote its migration would seem far below that 
necessary to promote the migration of an atom 
buried in the space lattice. Possibly the fact 
that our activation energies for lead, deduced 
from both Dr. Smith s data and the Hanff- 
stengel and Hanneman data, turned out to be 
approximately half of that found for the heat 
of self-diffusion, is quite significant with respect 
to the well-known fact that the finer the grain 
structure associated with a metal, the higher 
its creep rate at a given stress above the tem 
perature for recrystallization. 

W. H. BASSETT, JR.,* Hastings-on-Hudson, 
N. Y. The data obtained in our laboratories 
on common desilverized and copper lead, when 
using an entirely different test method from 
the one described by Dr. Smith, agree with the 
results reported by the author. 

The samples used in our laboratory consisted 
of is-in. lengths of extruded pipe approximately 
2% in. o-d. by y in. thick. The ends of the 
pipe were sealed by means of hollow plugs. 
The pipe was filled with oil, then hydraulic 
pressure was applied by means of a pump and 
accumulator, so that the samples were kept at 
constant pressure until failure occurred. By 
taking gauge measurements we were able to 
determine the rate of expansion of the tubes 
while determining the length of time required 
for rupture of the wall. Specimens were tested 
over a fairly wide range of time intervals and 
pressures. Rather high internal pressure caused 
failure in only 5 min. while at lower pressures 
the pipe lasted over 2 years before mpture 
occurred. 

Fig. lof shows a sample of pipe and the 
method of plugging the ends. Fig. n shows 
samples of burst pipe and indicates the type of 
fracture and the variation of expansion noted 
Some lead alloys expand to a considerable 
extent while others give very little expansion 

* Manager, Metallurgical Department, Anaconda 
Wire and Cable Co. 

t Figs. 10-13, 15, 17, 19, 21 were used in the paper 
by W. H. Bassett, Jr. and C. J. Snyder: Proc. Amer. 
Soc. Test. Mat. (1940) 40, 910-936. 



DISCUSSION 



173 



before bursting. The conditions reported by the 
author as to specimens apparently showing no 
creep for a considerable period of time and then 
making a sudden change in dimensions was 



formation of fine grains at the grain boundaries. 
Fig. 1 6 shows the structure of the same sample 
17 months after rupture. The fine grains are 
still noticeable at the grain boundaries. Fig. 17 




FIG. 10. SAMPLE OF PIPE WITH PLUGGED ENDS. 



also noted in the tests of the extruded pipe. 
Fig. 12 shows a group of typical curves drawn 
from the data obtained on various lead-pipe 
specimens. The samples used for this figure 
were common desilverized lead, chemical lead, 
copper lead and calcium lead. Two different 
specimens of copper lead are shown one 
sample from a tube and the other from a cable 
sheath. The extra working of the metal during 
the sheathing operation gives an entirely differ 
ent shape of curve and materially increases the 
life of the sample. Fig. 13 shows a summation 
of data obtained on various grades of lead and 
lead alloys as to the time required to burst the 
specimens when subjected to various internal 
pressures. Table i shows the composition of 
samples used in the preceding figures. Fig. 14 
shows a cross section of common lead pipe 
before test. The ring picture in the center 
shows a complete cross section and the enlarge 
ments show the structure of the top and bottom 
weld area and of the respective sides of the 
pipe. This pipe was subjected to 600 Ib. per 
sq. in. hoop stress and burst after 2 months. 
Fig. 15 shows the structure of the specimen 
one day after rupture occurred. Note the 



shows the structure of a copper lead pipe before 
test. This sample was subjected to 800 Ib. per 




FIG. ii. SAMPLES OF BURST PIPE. 



TABLE i. Chemical Analyses of Burst Test Specimens 
PER CENT 



Sample 
No. 


Lead 


Bismuth 


Copper 


Anti 
mony 


Tin 


Silver 


Tellu 
rium 


Nickel 


Calcium 


Mag 
nesium 


2264 


99.920 


0.065 


0.013 


0.0015 


o.ooo 












2261 


99.921 


0.066 


O.OII 


0.0015 


o.ooo 












2180 


99.880 


0.059 


0.059 


0.002 


o.ooo 


O.OOIO* 










2181 

2747 


99-884 


0.057 


0.057 


0.002 

(Similar 


0.000 

to Sampl 


s 2180 ai 


id 2181) 








2758 
2296 


99.826 
99.925 


O.IOO 
0.000 


0.072 
0.065 


O.OOI 
0.0003 


o.ooo 

0.000 


0.0015 
0.005* 


0.000 


0.000 
0.0044 






2445 
1228 
2856 


99.778 
99-905 
99-658 


o. 126 

0.000 

0.019 


0.059 
0.053 
0.005 


0.004 
0.0007 

0.002 


o.ooo 

0.000 

0.26 


0.013 
0.0062 


0.020 
0.000 


0.0044 


0.031 
0.025 


0.031 



* Nominal average* 



174 



CREEP AND RECRYSTALLIZATION OF LEAD 



26 
24 



20 



JI6 

14 

. 10 
S 8 

8 A 

b 6 

~ 4 
2 



" Sample 4 /One/ of Lead Stress, psi. 

No. 1228- 0.03 per cent Calcium I50Q 

~ No. 2/80- AS. f M. Grade M plus 

0. 06 per cent Copper (Tube] - BOO 

-No,22S4-A.$.T.M.Gradem BOO 

No.2296-A.SJ.M. Grade U 800 

. No.2747-A.SXM. Grade IHp/as 

0.06 per cent Copper (Sheath)- 




8000 16000 24000 32000 40000 48000 56000 64000 

Time , min 
FIG. 12. TYPICAL CURVES FROM DATA ON LEAD-PIPE SPECIMENS. 



JO 000 000 

1000000 
year- 

JOOOOO 




10 000 
week 



1000 



fOO 
hour- 



- 



<ft 



Comparison of Curves from Fig. 1\ - 
with Other Investigators Resutf$ ~ 

Basse ft Snyder 

Moore. Beffy 8c Doll ins - 

H.S.Phelps E 



i 




400 800 1200 1600 2000 2400 
Hoop Stress, psi 

FIG. 13. DATA REGARDING TIME REQUIRED TO BURST SPECIMENS SUBJECTED TO INTERNAL PRESSURE. 



DISCUSSION 



ITS 




16 



FIG 14. CROSS SECTION OF COMMON LEAD FIG. 17. STRUCTURE OF COMMON LEAD PIPE 

4 " 



BEFORE TEST. 



pIpE . 

FIG is -STRUCTURE OF "SPECIMEN OF FIG. 14 FIG. 18. SAMPLE OF FIG. 17 TEN WEEKS AFTER 

RS. 



ONE D4Y AFTER RUPTURE. 



SPECIMEN BURST. 



ONE D4Y AFTER RUPTURE. . 

FIG 16 -STRUCTURE OF SAME SAMPLE SEVEN- FIG. IQ.-SAME SAMPLE TWENTY-SIX MONTHS 

" 



TEEN MONTHS AFTER RUPTURE. 



ATTER SPECIMEN BURST. 



176 



CREEP AND RECRYSTALLIZATION OF LEAD 




FIG. 20. STRUCTURE OF ORIGINAL COPPER- 
LEAD SAMPLE BEFORE TEST. X IOO. 

FIG, 2 1. SAMPLE OF FIG. 20 TEN WEEKS AFTER 

FAILURE. X 100. 

FIG. 22, SAME SPECIMEN TWENTY-SIX MONTHS 
AFTER FRACTURE. X IOO. 



sq. in. hoop stress and burst after 2 months. 
Fig. 1 8 shows the structure of the sample 
10 weeks after the specimen burst. There is a 
marked change in the crystal structure. Fig. 19 
shows the same specimen 26 months after the 
specimen burst. Fig. 20 shows the structure of 
the original copper-lead sample before test, at 
a magnification of 100 diameters. Fig. 21 shows 
this copper-lead sample 10 weeks after failure. 
At loo diameters the small crystals at the grain 
boundaries and the irregular shape of the grain 
boundaries is readily noted. Fig. 22 shows the 
same specimen 26 months after fracture oc 
curred. The small crystals can still be noted 
between the grain boundaries and the wavy 
appearance of the crystal grain boundaries is 
very noticeable. 

The similarity of results obtained by the 
author on high-purity lead with flat strips and 
the results we obtained with extruded pipe 
from the commercial grades of lead would seem 
to indicate that the author s conclusions can be 
expanded to cover a broader range of material. 

E. E. SCHUMACHER,* New York, N. Y. The 
irregular shape of the time-elongation curves 
as reported here and by others for the speci 
mens tested at low stress is extremely inter 
esting. There seems to be general agreement 
among those who have published data on creep 
that this irregularity is most pronounced in low 
stress tests on lead of quite high purity and in 
specimens having large grain size. Many 
investigators have observed that extension of 
lead alloys under loads causing creep may take 
place by grain deformation, by grain-boundary 
separation which leaves voids in the material, 
or by motions of the grains relative to each 
other along the boundaries. 

A possible explanation for the irregular creep 
rates reported by Dr. Smith may be given for 
those instances where the grain size is large and 
the stress is low. Under these conditions it is 
probable that the last-mentioned mechanism 
of extension is dominant. In a sample contain 
ing a random grouping of grains of various 
shapes, sizes and orientations it is reasonable 
to suppose that some configurations are much 
more favorable than others to motion along 
the grain boundaries. After a certain amount of 
motion there will have occurred some rotation 
until irregularities interlock and prevent further 
motion. Additional extension of the specimen 

* Bell Telephone Laboratories. 



DISCUSSION 



177 



would then have to take place by slip within 
the grains themselves, which, at the low 
stresses employed, would take place slowly. 
This mechanism, after a period of time, would 
once more bring some grains by rotation into 
positions favorable to intergranular motion and 
again permit extension of the specimen. 

This explanation of the irregular creep rate 
is consistent with the observation of the author 
that coarse-grained materials flow at more 
irregular rates at low creep stresses than fine 
grained materials. In the latter case, individual 
instances of unfavorable orientation between 
grain neighbors would have a less pronounced 
effect on the creep rate of a specimen as a 
whole. 

R. L. TEMPLIN,* New Kensington, Pa. I 
realize the space limitations imposed on authors 
of a paper such as this, yet, in view of our own 
experience in conducting creep tests and the 
brief description given by the author of his 
apparatus and test procedure, I would like to 
ask what precautions were taken to make sure 
that his creep tests were free from vibration 
effects. The schematic drawing of the specimen 
assembly shown in Fig. i does not clearly 
indicate the details of the support used for each 
specimen. While the author states that "no 
vibration was detectable under the most severe 
conditions of shock in the adjacent surround 
ings," no details are given of how he assured 
himself of this fact. Under conditions of insta 
bility that must occur during a creep test, it 
would seem logical to expect that vibration 
might under suitable conditions cause irregu 
larities in the creep curves similar to some of 
those shown in the data presented by the 
author. 

B. B. BETTY, f Huntington, W. Va. In 
connection with the question of possible effects 
of vibration on rates of creep, two facts offer 
considerable comfort to the laboratory worker. 
In a cooperative investigation to test the 
reproducibility of creep rates on K-2o steel, a 
number of different laboratories in different 
parts of the country obtained almost identical 
creep curves. It seems safe to presume that 
they were subjected to different intensities of 
vibration and hence ordinary vibration result 
ing from street traffic probably has little 

* Chief Engineer of Tests, Aluminum Company of 
America. 

t Research Laboratory, International Nickel Co. 



deleterious effect on creep tests. Secondly, in a 
single laboratory, specimens in adjacent creep 
frames may have pulsating creep rates whose 
undulations are not in phase; i.e., their maxi 
mum rates do not occur simultaneously. The 
argument often advanced to explain this 
pulsating creep rate phenomenon, that it is due 
to vibrations, does not seem to be sufficient. 
If the cause of this cyclical effect is to be 
attributed to shock, a second assumption is 
necessary; namely, that the metal varies from 
time to time in its responsiveness to shock 
waves. 

A. A. SMITH, JR. (author s reply). Although 
the energy value deduced from the data falls 
short of the heat of self -diffusion for lead, there 
still may be some relationship between creep 
and diffusion as suggested by Dr. Kanter. 
However, the relationship is probably very 
complex and such factors as grain size, rate of 
deformation and recovery must be considered. 
As the paper indicates, if the rate of creep is 
low the type of deformation produced does not 
cause recry stabilization; that is, recovery 
proceeds at such a rate and in such a manner as 
to offset the lattice distortion. This recovery 
process probably is due to some type of atomic 
movement, but whether it may be of the same 
nature as self-diffusion is not entirely clear 
from our present knowledge. 

The data given by Mr. Bassett on burst tests 
of various brands of lead tend to confirm the 
author s findings that in long-time tests no 
recrystallization can be found. 

Mr. Schumacher has put forth a very inter 
esting suggestion regarding the irregular shape 
of the time-elongation curves and in some 
instances it may explain the phenomenon. 
However, when there are long periods of zero 
extension, it is difficult to understand how the 
grains can be rotated into more favorable 
positions for further extension. It is conceivable 
that during these waiting periods certain 
obstructions to further motion are relieved 
by the thermal vibration of the atoms. This 
idea appears consistent with the data relative 
to the lack of recrystallization at low creep 
rates, which probably is due to some recovery 
effects in the specimens. 

Undoubtedly, the presence or absence of 
vibration during creep testing is an important 
factor, as Mr. Templin suggests and under 



i 7 8 



CREEP AND RECRYSTALLIZATION OF LEAD 



suitable conditions might cause irregularities 
in the creep curves. In the same creep room 
where lead is being tested there are copper 
alloy wires of 6o-in. gauge length under test 
and many hours spent observing these speci 
mens with a microscope have failed to show 
any vibration despite the high accuracy of 
measurement. Moreover, as mentioned in the 
paper it is evident that a sudden shock should 



affect in the same manner two adjacent 
specimens suspended from the same angle 
support, nor can this explain long periods of 
zero extension. 

Other laboratories have also noted this type 
of creep curves and Mr. Betty s remarks 
confirm the author s belief that vibration does 
not appear to be sufficient to explain the 
pulsating creep phenomenon. 



Tensile Properties of Rolled Magnesium Alloys Binary Alloys 
with Calcium, Cerium, Gallium, and Thorium 



BY JOHN C. MCDONALD,* 

(Cleveland Meeti 

THIS report is a continuation of an earlier 
one with a similar title, 1 to which the reader 
is referred for such details of procedure as 
do not appear here. A brief summary will 
be given of the objects and methods of the 
research at this time. 

It was pointed out in the previous paper 
that properties of magnesium alloys, as re- 
ported by different workers, often differ. 
This was thought to be due to differences 
in the preparation of the specimens tested. 
The ideal way to compare alloys would be 
to prepare each of them in so many ways 
that reasonable assurance would exist that 
the best properties of each alloy had been 
developed. The alloys would then be com 
pared on the basis of these properties. 

The treatment actually given the alloys 
of these studies is necessarily a compromise 
with completeness. Nevertheless, enough 
treatments have been given so that a con 
siderable degree of assurance exists that 
the properties obtained are representative 
of the alloy as such, and not merely of the 
manner in which it was worked and 
heat-treated. 

PROCEDURE 

The alloys were cast in the form of 
cylindrical ingots, which were extruded into 
bar before being rolled. Some of the ingots 
were made in 2 -in. diameter molds whereas 
all of the alloys of the previous paper were 
cast in 4-in. molds. For the larger ingots, 



Manuscript received at the office of the Institute 
July 15, 1940. Issued as T.P- 1247 in METALS TECH 
NOLOGY, December 1940. 

* The Dow Chemical Co., Midland, Michigan. 

1 References are at the end of the paper* 



JUNIOR MEMBER A.I.M.E. 

ng, October 1940) 

the extruded bar was 2 by 0.140 in.; for 
the 2-in. ingots, it was i by o.ioo in. 
Preliminary work indicated that strip 
rolled from both sizes of the bars would 
have substantially the same properties. 

The bar was rolled at some temperature 
in the range 400 to 8ooF. ? for a total 
reduction of 50 per cent. The temperature 
and reduction per pass were so selected 
that no cold cracking occurred, but were 
such as to leave the metal in a semihard 
condition. The hot-rolled metal was an 
nealed at several temperatures in the range 
400 to 8ooF.; it was also cold-rolled, and 
given solution and aging treatments. Two 
standard test bars corresponding to each 
treatment were pulled in tension, and the 
average was obtained. Twelve treatments, 
in all, were given each alloy. 

The properties of the annealed specimen 
having the highest percentage of elongation 
were selected to represent the alloy in the 
annealed condition. The properties of the 
cold-rolled specimen having the highest 
yield strength were selected to represent 
the alloy in the cold-rolled condition. 
These properties were then plotted against 
percentage composition by analysis. Smooth 
curves were drawn among the points, thus 
graphically averaging the results over the 
entire range studied. The percentage of 
elongation in the cold-rolled state was not 
plotted, since it was always low, and varied 
in an irregular manner with composition. 

ALLOYS STUDIED 

Magnesium-calcium Alloys. Fig. i pre 
sents the results on the alloys of the mag- 



179 



i8o 



TENSILE PROPERTIES OF ROLLED MAGNESIUM ALLOYS 



nesium-calcium system. There is a rapid 
rise in ductility up to o.i per cent Ca, 
which is maintained to 0.3 per cent Ca, 
after which it slowly decreases. The rise 



These alloys were made with sublimed 
magnesium, which accounts for the lower 
strength at o per cent Ga in Fig. 3, as 
compared to Figs, i and 2. The rate of 




FIG. i. -RESULTS ON MAGNESIUM-CALCIUM 

ALLOYS. 

in strength is most rapid up to 0.3 per 
cent. 

Magnesium- cerium Alloys. The results 
for the magnesium-cerium alloys are em 
bodied in Fig. 2. We see that cerium be 
haves much like calcium. There is a rapid 
rise in ductility up to about o.i per cent, 
after which it slowly decreases. The yield 
strength in the annealed condition con 
tinues to increase out to the limit of com 
position studied. The rise in strength is 
most rapid up to o.i per cent. The prop 
erties of the cold-rolled metal above 0.5 per 
cent could not be obtained, because the 
hot-rolled strip was too hard to withstand 
further cold-rolling. These alloys were 
made by adding 98 to 99 per cent pure 
cerium ? free from other rare earths; not the 
"Mischmetal" containing 50 per cent Ce, 
which often is referred to as "cerium." 

Magnesium-gallium Alloys. Fig. 3 shows 
how the expensive metal gallium increases 
the ductility of magnesium. Because of the 
nature of the constitutional diagram ? and 
the similarity of gallium to aluminum, it is 
probable that the ductility would eventually 
decrease, as the percentage of gallium 
increased. 



f 

|- 

r 

**J 

i 
I 20 

VI I 

I 

to 

(. 
















., ^ 


i -i 


- 


.^-- 









f 












-+ 


r*--^ 
i^ 


, 

<-. c 


\ 


,-+ *" 


.---" 






X 

^N 


\ 














a 


^ 


\ 






I 


, COLD KfOLLCO " 
, AMNZ.ALE.D 
. COL0 ROLLZD * 
, ANNkL0 
r j ANN&ALZO 


r 
^ 

a 


-s 


N, 



FIG, 2. RESULTS ON MAGNESIUM-CERIUM 
ALLOYS. 

increase of strength is most rapid up to 
o.i per cent Ga. 

Magnesium-thorium Alloys. The effect 
of thorium is shown in Fig. 4. The base 
magnesium had been sublimed. There is as 
yet some uncertainty as to the precise rate 
of rise of the elongation in the annealed 
state, which will require further work to 
eliminate. The existence of ductile alloys 
at 2 to 3 per cent Th is a positive finding, 
however. Age-hardening is observed at 
2 per cent and above. 

DISCUSSION OF RESULTS 

Table i summarizes the properties of 
the best alloys of each system, within the 
limits studied, considered from the stand 
point of ductility. 

These metals belong to the group of 
alloying elements that increase the ductility 
of magnesium. The magnitude of the in 
crease in ductility is such as to put them 
into a class with zinc, silver, aluminum, 
cadmium and thallium. 1 

The constitutional diagram of all the 
binary systems reported on to date is 
known, except for thorium. 2 3 A study of 
these diagrams leads one to the conclusion 



JOHN c. MCDONALD 



181 



that the maximum of ductility exists at a 
composition where the added element is 
largely in solid solution. Presence of a 
second phase decreases the ductility. The 




T5.C& 

rs. Ah 
Y,ca 
YS.AA 



FIG. 3. RESULTS ON MAGNESIUM-GALLIUM 
ALLOYS. 

converse of this first proposition is not true; 
the existence of a range of solid solubility 
does not guarantee the production of an 
increase in ductility (tin, lead, bismuth). 
Since thorium does produce an increase in 

TABLE i. Properties of Alloys of Highest 
Ductility 





Elonga 


Yield 


Tensile 


Composi 


tion, 
Per Cent 


Strength, 1000 
Lb. per Sq. In. 


Strength, 1000 
Lb. per Sq. In. 


tion, 








Per Cent 














Annealed 


Annealed 


Cold- 
rolled 


Annealed 


Cold- 
rolled 


0.3 Ca 


17 


20 


29 


33 


37 


0.2 Ce 


19 


18 


25 


34 


34 


0.8 Ga a . .. 


21 


19 


29 


33 


37 


2.0 Th... 


22 


IS 


26 


33 


32 



Sublimed Mg base. 

ductility, and also causes age-hardening, 
it is probable that there is a considerable 
range of solid solubility in the magnesium- 
thorium system. It has been observed also 
that in systems where the range of solid 
solubility is very small, there is little or 
no increase in ductility (copper, nickel, 
antimony). This is in agreement with the 
previous statement about the effect of the 
presence of a second phase. 



The ductility of single crystals of mag 
nesium, magnesium plus aluminum and 
magnesium plus zinc has been studied. 2 
Both aluminum and zinc cause the ductility 




FIG. 4. RESULTS ON MAGNESIUM-THORIUM 
ALLOYS. 

of a single crystal to decrease. The increase 
in ductility of the polycrystalline metal 
may be due to modes of deformation that 
are not observed in single crystals. Cer 
tainly the crystallites are subjected to a 
much more complicated system of stress 
than can be enforced on a single crystal. 
It is probable that the other ductility- 
producing elements act in the same way. 
In both single and poly-crystals, foreign 
atoms increase the strength. 

CONCLUSION 

Calcium, cerium, gallium, and thorium 
increase the ductility of magnesium mark 
edly, and also increase its strength. In this 
respect they are similar to zinc, silver, 
aluminum, thallium, and cadmium. They 
are much different from antimony, bis 
muth, copper, lead, nickel, and tin, which 
have little or no beneficial effect on ductility. 

REFERENCES 

1. J. C. McDonald: Tensile Properties of Rolled 

Magnesium Alloys, I. Binary Alloys with Al, 
Sb, Bi, Cd, Cu, Pb, Ni, Ag, Tl, Sn and Zn. 
Trans. A.I.M.E. (1940) 137, 430, 

2. A. Beck: Magnesium und Seine Legierungen. 

Berlin, 1939. J. Springer. 

3. J. L. Haughton and W. E. PrytherclK Magnesium 

and Its Alloys. London, 1937. His Majesty s 
Stationery Office. 



182 



TENSILE PROPERTIES OF ROLLED MAGNESIUM! ALLOYS 



DISCUSSION 

(L. W. Eastwood presiding) 
G. EDMUNDS,* Palmerton, Pa. The effects 
upon ductility of either aluminum or zinc 
present in solid solution in magnesium are 
given as an increase of ductility for rolled, 
annealed specimens, but a decrease of ductility 
for single crystals. This difference might well 
be the result of a change of mode of deforma 
tion, as suggested by Mr. McDonald. One can 
ask what is meant here by the expression 
"mode of deformation." In a broad sense it is 
inclusive. Deformation can occur as a dis 
placement of grains as a whole with a minimum 
of grain distortion, or primarily as grain 
distortion with a minimum of relative grain 
displacement. In the first case, a relatively 
brittle single crystal and a ductile polycrystal- 
line sample of the same alloy would be con 
sistent, and strengthening, as by alloyage, of 
the grain could contribute to increased ductility. 
Should the deformation consist primarily of 
grain distortion, the complex stress system set 
up in polycrystalline aggregates may become 
especially important. Then, however, in 
randomly oriented specimens there would be 
present crystallites most unfavorably oriented 
for deformation and miniature fractures would 
be anticipated unless the stress requirement for 
fracture exceeded that which could be imposed 
by neighboring crystallites more favorably 
oriented for deformation. If, on the latter 
account, fractures were hindered or precluded, 
the aggregate might exhibit good ductility 
while a single crystal of unfavorable orientation 
might be brittle. Grain refinement, as by an 
alloying addition, even though it tended to 
embrittle single crystals, might then toughen 
rolled, annealed aggregates. On the other hand, 
if the miniature fractures should develop and 
propagate, owing to stress concentrations 
caused by them, the polycrystalline material 
might be nearly as brittle as would single 
* Research Division, The New Jersey Zinc Co. 



crystals that were oriented unfavorably for 
deformation. 

Grain orientation may also be significant in 
itself, and since rolled specimens even after 
annealing commonly exhibit a special orienta 
tion texture, which may be either favorable or 
unfavorable for a particular deformation, 
ductility may be obtained with such specimens 
even though single crystals deformed in some 
ways may be brittle. Alloying, by altering high- 
temperature modes of deformation, may result 
in a different orientation texture. 

It has not been intended here to give a com 
plete analysis of the possible ways of rational 
izing an embrittlement by alloying of a single 
crystal with a toughening by the same alloying 
of rolled or rolled and annealed specimens. 
Obviously many other factors, such as grain 
shape, strain-hardening, mechanisms of slip 
and twinning, and dendritic structures could 
be brought into the discussion. Various con 
siderations that have been given seem to suffice 
to show, however, that it does not seem strange 
that there may exist a reversal in effects of 
alloying upon the ductility of single crystals 
and polycrystalline aggregates. 

J. C. MCDONALD (author s reply) .Mr. 
Edmunds remarks throw considerable light 
on the experimental findings. By "modes of 
deformation 1 is meant the different ways in 
which a single crystal can change its shape, 
under stress, without fracturing; i.e., slip or 
twinning on a certain plane, or planes. Only a 
few such mechanisms have been observed in 
isolated single crystals of magnesium and a few 
of its alloys. In polycrystalline material, the 
interaction of the crystals may bring other 
mechanisms into play. As Mr. Edmunds points 
out, these new mechanisms may be different 
for alloys than for the pure metal. This would 
create a difference in the preferred orientation 
of the rolled metal, in the first place, and, in the 
second place, would cause it to deform differ 
ently when tested. 



Grain Orientation of Cast Polycrystalline Zinc, Cadmium and 

Magnesium 



BY GERALD EDMUNDS,* MEMBER A.I.M.E. 

(Cleveland Meeting, October 1940) 



CASTINGS of pure metals and many alloys 
usually have a coarse-grained structure 
characterized by long columnar grains 
throughout the main body of the casting. 
Frequently, the surface exhibits finer, some 
times nearly equiaxed grains, and such 
grains may also compose other parts of the 
casting. This paper deals primarily with 
determinations of the grain-orientation 
textures of the columnar and surface grains 
of poly crystalline zinc castings. It includes 
some results upon a cadmium casting and a 
magnesium casting. A hypothesis is devel 
oped to account for the observations on 
columnar grains and to use as a basis for 
predicting orientations in other cast metals. 

A general description of the position of 
the columnar grains in castings is that their 
long axes are approximately parallel to the 
direction of the thermal gradient during 
crystallization. Thus, in cylindrical castings 
the columnar grain axes tend to be radial, 
and in large flat castings where most of the 
cooling is through one or both of the large 
mold surfaces they tend to be perpendicular 
to the mold surfaces. At the edges and 
corners the direction of heat flow and there 
fore the position of the grains is more com 
plex. Actually observation shows that even 
in castings made in flat molds the cooling 
seems to be somewhat irregular, since 
columnar grains tend to emanate from 
certain areas of the cooling surface, indicat 
ing that thermal contact was better there 
than in neighboring areas. 

Manuscript received at the office of the Institute 
June 29, 1940. Issued as T.P. 1244 in METALS 
TECHNOLOGY, October 1940. 

* Investigator, Research Division, The New Jersey 
Zinc Co. (of Pa.}, Palmerton, Pa. 



The nucleus from which a columnar grain 
grows may be so oriented that its direction 
of maximum growth rate does not coincide 
with the direction of the thermal gradient. 
In this case the long axis of the grain prob 
ably assumes an intermediate position. 

Notwithstanding these minor irregulari 
ties, the situation does exist that the long 
axes of the columnar grains are generally 
disposed approximately perpendicular to 
the mold surface. 

Schmid and Wasserman 1 and Nix and 
Schmid 2 have determined the orientation 
texture of columnar grains in castings of 
several cubic, tetragonal, hexagonal close- 
packed and rhombohedral metals. For zinc 
and cadmium, they report the basal plane 
to be parallel to the columnar grain axes, 
with no preference for any particular direc 
tion in this plane to coincide with the grain 
axes. This texture, termed a ring fiber tex 
ture, 3 is uncommon in metals; zinc and 
cadmium are the only ones reported to 
exhibit it. Generally, it is stated, a simpler 
texture exists in which a principal, though 
not necessarily the most closely packed, 
crystallographic direction is parallel to the 
direction of heat flow. In cast magnesium 
(also hexagonal close packed) they report 
a digonal axis I, [100] direction, parallel to 
the direction of heat flow. The writer is not 
aware of any other orientation-texture 
determinations on cast polycrystalline 
hexagonal metals. 

In view of the observations reported in 
the present paper it is of interest also to 
note that Nix and Schmid stated that a 



183 



1 References are at the end of the paper. 



1 84 



GRAIN ORIENTATION OF CAST ZINC, CADMIUM AND MAGNESIUM 



departure from randomness of orientation 
is especially to be expected in the columnar 
grains; they do not seem to have deter 
mined the orientation texture at the surface 



tion adopted for most of the specimens con 
sisted in sawing with a power saw to within 
about % in. of the surface to be examined, 
then milling with a cut of about Ke i n - 



DIRECT BEAM 
/SPECIMEN REMOVED) 




FIG. i. TYPICAL DIFFRACTION PATTERN OBTAINED FROM COARSE-GRAINED ZINC SPECIMEN 

(FlG. 7) USING ROTATING-TRANSLATING SPECIMEN X-RAY SURFACE-REFLECTION METHOD OF 
ORIENTATION-TEXTURE DETERMINATION. IMPORTANT REFLECTIONS ARE INDEXED ON FIGURE, 

of castings. Desch 4 states that surface chill 
crystals appear to be randomly oriented. 

EXPERIMENTAL PROCEDURE 

Thirty-six zinc castings were prepared 
and their orientation textures determined 
at various levels. The castings were made in 
a variety of sizes, using cast iron, steel, and 
graphite molds and a wide range of metal 
and mold temperatures and other condi 
tions. Only a few representative ones will be 
described. One casting each of cadmium 
and magnesium was studied. 

Methods Used for Orientation 

Determinations* 

Orientation textures were determined at 

the original surfaces and at various levels in 
the castings on surfaces parallel to the 
cooling surface. The internal surfaces re 
quired great care in preparation to avoid 

mechanical twinning, which would lead to 
erroneous results. The method of prepara- 



*See Indexing of Crystallographic Features, page 
185. 



followed by successively lighter cuts until 
the last, about 0.002 to 0.005 i n -> brought 
the specimen to about ^4 in. from the final 
surface. The remaining metal was removed 
by metallographic polishing and deep etch 
ing, usually with a solution of 20 grams 
CrOs, 1.5 grams Na 2 S04 and 100 c.c. water. 
Several methods of determining orienta 
tions were used; the most significant and 
the only ones reported here were by X-ray 
diffraction. The principal method was of 
the surface-reflection type, using a tech 
nique that has been designated as the 
rotating-translating specimen X-ray sur 
face-reflection method. The specimen is 
rotated about the fiber axis, in this case the 
direction perpendicular to the flat mold 
surface, and simultaneously moved side to 
side and back and forth in the plane of the 
specimen surface with the collimated X-ray 
beam always incident at the same angle to 
the specimen surface. The diffraction pat 
tern is recorded by a photographic film 
beyond the specimen and perpendicular to 
the incident beam. For these tests Mo K a 
radiation incident upon the surface at 
about 8, and a specimen to film distance of 



GERALD EDMUNDS 
-EsL 



-a, 




FIG A STEREOGRAPHIC PROJECTION OF THE ZINC CRYSTAL (AXIAL RATIO 1.8563 AT 27C.). 

Showing the crystal planes (JWM) of high atomic density, certain important directions [hkl], 
the basal and prism planes (hkl}T of one twin of the first generation all referred to two equal 
horizontal axes! * and ^ intersecting at an angle of 120, and a third, vertical axis c at jight 
angles to themj and the positions of the Miller hexagonal-rhombohedral axes with their Miller 
indices [hkl]M. 

Indexing of Cry statto graphic Features: The three-index system has been used throughout. 
This is based upon two equal axes, 01 and * 2 , along close-packed directions and intersecting at 
an angle of 120 and a third axis c normal to the plane of the other two. A stenographic projection 
of the zinc crystal is given here (Fig. A). The projections of other hexagonal closepacked metals 
are similar, differing only in the angle between the hexagonal axis and the pyramid planes (because 
the axial ratio (c/a) differs). . 

The four-index system (MiUer-Bravais) indices for planes are obtained by supplying between 
the second and third index an additional one equal to the sum of the first two with the sign 
reversed, thus: 

(210) in the three-index system ss (2110) in the Miller-Bravais system. 

Transformations of indices of crystallographic directions are not made in this way. (Line 
indices are the coordinate differences between adjacent atom positions along the crystallographic 
line (or direction).) In contrast to the isometric system (cubic crystals), crystallographic directions 
normal to crystallographic planes do not in general bear the same indices as the planes. The 
normals to pyramidal planes have no crystallographic significance. 



j 86 



GRAIN ORIENTATION OF CAST ZINC, CADMIUM AND MAGNESIUM 



7 cm. have been used. The area of specimen 
surface examined was about one square 
inch. With this method, every grain extend 
ing to the surface yields diffraction effects 




rotated about its axis and moved back and 
forth in the direction of the axis. 

Back-reflection Laue X-ray patterns 
were prepared from 20 grains of one speci- 

Top surface. Too uneven for orien 
tation determination. 

"About Jfg in, below top surface: 
(ooi) 9, moderate;* (100) 
over wide angular range, weak. 

in. above bottom surface: (100) 
3, very strong; (100) 20, 
weak. 



in. above bottom surface: (100) 
13 intensity maximum at 
parallel position; (oor) + 10. 



.120 in. above bottom surface: 
(101) 9, strong; (100) 20, 
moderate; (ooi), a single, weak 
reflection. 



0.052 in. above bottom surface: 
(101) 6; (ooi) 8; (100) 
over wide angular range, weak. 

Original bottom surface: (ooi) 



FIG. 2. CASTING No. i. HALF OP VERTICAL CROSS SECTION. 
Etch: concentrated HO. Original magnification 2; reduced Yi in reproduction. 
The orientations* at various levels as indicated are given at the right; no determinations were 
made beyond the ends of the ruled lines. The ruled lines are the traces of the planes of polish upon 
which orientations were determined. 

* Where orientations are expressed, for example, as: "(ooi) 9, moderate," this is to be read as: 
"The (opi) planes have a mean position parallel to the plane of polish, the angular variation from 
this position is 9." The plane of polish upon which the orientation is determined is perpendicular 
to the average direction of heat flow; correspondingly, it is approximately perpendicular to the axes 
of the columnar grains. The more predominant orientations are given first. This means of express 
ing orientation textures has been used throughout. 



at some time during the exposure, so that 
even coarse-grained specimens yield a pat 
tern that is readily interpreted. Fig. i is a 
typical diffraction pattern. The validity of 
this method was established by comparison 
of results with those obtained on individual 
grains by the back-reflection X-ray method. 
For cylindrical specimens having radial 
grains the X-ray beam is incident tangen- 
tially on the specimen. The specimen is 



men from casting No. 2 (zinc) and 10 
grains of casting No. 7 (magnesium). The 
individual grains of the stationary specimen 
were irradiated with a coIUmated beam of 
unfiltered copper X-rays (50 kv. P) incident 
upon the specimen normal to the cooling 
surface. The pattern was recorded on a flat 
photographic film at a distance of 3 cm. 
The orientation was determined by meas 
urements on stereographic projections. 



GERALD EDMUNDS 



I8 7 



DESCRIPTION OF SPECIMENS AND 
RESULTS OBTAINED 

Casting No. i. Special High Grade zinc* 
was poured into a heavy, open-top cast-iron 
B 



Orientations at various levels are given 
beside a macrograph of a typical cross sec 
tion in Fig. 2. 

Casting No. 2. High Grade zinc was 
poured 2% in, deep into a large cast-iron 

A 




and 15 
jr la 3 VERTICAL SECTIONS, iH INCHES WIDE, THROUGH 2%-iNCH DEEP CASTING No, 2, SHOWING 

TWO SIDES OF SPECIMEN LATER. CUT FOR EXAMINATIONS OF HORIZONTAL SECTIONS. 

Numbers at right indicate levels at which macrostructures and orientation textures 
(Figs. 4 to 15) were determined, and correspond to those figure numbers. Letters A and B above 
the macrographs are used to show the correspondence of sides as designated in the captions to 
Figs. 4 to 15. Etch: Concentrated HC1. 

Original magnification 1.8; reduced l / in reproduction. 



mold (about 40 in, long by 3 in. wide by 
2 in. deep). Cooling proceeded chiefly from 
the mold, and partly from the exposed top 
surface. 



* Grades of zinc used are designated according to 
the Standard Specification for Slab Zinc (Spelter); 
A.S.T.M. designation. B6-37. 



tray. Cooling from the top surface was 
prevented by using a heater above the 
casting so that the cooling was from the 
bottom. 

One specimen, iK in square by 2% in. 
high, was cut far enough from any edge of 
the casting to avoid edge cooling effects. 



i88 



GRAIN ORIENTATION OF CAST ZINC, CADMIUM AND MAGNESIUM 



The macrostructure of vertical and hori- 
zontal sections of the casting and the 
orientation-texture data are shown in Figs, 
3 to 15. 

The orientations of 20 individual grains, 



and that freezing from the top was not 
prevented but was accentuated by blowing 
cold air over the surface. 

A specimen, 2 in. square by 1% e i n - high, 
was cut from near the center. The macro- 




FIG. 4: original top surf ace: (ioo) 

8, strong; (ooi) 8, weak. 
D FiG. 5: He i n - below top surface: 
(ioo) 9. 



Fig. 6: y^ in. below top surface: 

(IOO) 12. 

Fig. 7: 2 in. above bottom surface: 
(ioo) 13. 



Fig. 8: i in. above bottom surface: 

(ioo) 15, very strong ;(oo i ) 

10, weak.* 
Fig. 9: }4 in. above bottom surface: 

(ioo) 19, strong; (ooi) 19, 

moderate.* 



FIGS. 4 TO 9. HORIZONTAL SECTIONS THROUGH 2%-iNCK DEEP CASTING No. 2. 
Etch, except Figs. 4 and 15, which were not etched: concentrated HC1. X 1.8. 
Only about a quarter of each surface is shown; the top edge is adjacent to side A and the left 
edge to side B of Fig. 3. .... 

Orientation textures are given at the right. The means used in describing orientations is given 
in the explanatory note under Fig. 2. 

* The (ooi) plane reflections probably derived from mechanical twins produced during prepara 
tion of the specimen. Other determinations (Figs. 5 and 7) likewise yielded these originally, but 
after careful repolishing and etching, which removed many of the surface twins, no (ooi) plane 
reflections were obtained. 
f A poorly defined pattern. 

Fig. 16, at the level of Fig. 7, 2 in. above the 
bottom surface, were also determined by 
the back-reflection Laue method and are 
shown stereographically in Fig. 17. 

Casting No, 3. Casting No. 3 was like 



No. 2 except that it was only 



. thick 



structure on a vertical section and the 
orientation texture determined on hori 
zontal sections at four levels are given by 
Fig. 18. 

Casting No, 4. High Grade zinc was 
poured into a g-in. long by s-in. wide by 



GERALD EDMUNDS 



189 



7-in. deep mold. One end of the mold was 
of water-cooled steel, and the other end, 
sides and bottom were (preheated) Tran- 
site, and the top was open. 
Specimens were cut through the grains 




diameter vent-end grip section of this bar. 
A section through the bar and the orienta 
tion data for various levels in it are shown 
in Fig. 20. The removal of metal to the 
levels examined was entirely by etching. 



Fig. 10 : y in. above bottom surface: 
(100) 14. 

Fig. ii : 5^ 2 in. above bottom sur 
face: (100) 29. 



Fig. 12: Jf e in. above bottom sur 
face: (100) over a wide angular 
rangejf (ooi) over a wide angular 
range, f 

Fig. 13: J^ 2 m - above bottom sur 
face: (ooi) 4, moderately 
strong; (100) 31, moderate. 



Fig. 14: bottom surf ace: (ooi) 9, 

very strong; (100), extremely 

weak. 
Fig. 15: original bottom surface: 

(ooi) 7, very strong; (100), 

extremely weak. 



FIGS. 10 TO 15. CAPTION AND NOTES UNDER FIGS. 4 TO 9. 



grown from both the steel end plate and 
one of the Transite sides. 

A horizontal section through the entire 
casting is shown in Fig. 19. Orientation 
textures were determined on vertical sec 
tions as indicated. 

Casting No. 5. Special High Grade zinc 
was cold-chamber die-cast in the form of a 
standard* J-in. diameter round tensile 
bar. A specimen was cut from the %-in. 



* Tentative Specifications for Zinc-base Alloy Die 
Castings, A.S.T.M. Designation, B86-38T. 



Casting No. 6, Cadmium (99.9+ per 
cent Cd grade) was cast about iK in. deep 
in a 4-in. square Transite-sided mold hav 
ing a 2-in. thick steel bottom. A specimen, 
ij^ in. square by i.i in. high, was cut from 
the center of the casting. 

The macrostructures of a horizontal sec 
tion, i.i in. from the bottom of the casting, 
and a vertical section showing the structure 
from the bottom upward for i.i in., to 
gether with orientation data, are given in 
Fig. 21. 



190 



GRAIN ORIENTATION OF CAST ZINC, CADMIUM AND MAGNESIUM 



Casting No. 7. Magnesium* cast about 
6-in. diameter by 20 in. high with cooling 
entirely from the bottom. The (horizontal) 
section used was cut about 15 in. from the 




FIG. 1 6. HORIZONTAL SECTION THROUGH 
2 Jg-INCH BEEP CASTING NO. 2, 2 INCHES ABOVE 
BOTTOM SURFACE. 

Etch: concentrate HCL Numbers designate 
grains whose individual orientations are given 
by Fig. 17. Original magnification about 1.8; 
reduced J4 in reproduction. 

bottom of the casting. The surface was 
prepared by lathe turning followed by deep 
etching. Because of the large grain size the 
only orientation determination on this cast 
ing was by the back-reflection Laue tech 
nique. The orientation of 10 grains was 
determined, as indicated by Figs. 22 and 23. 

Results 

These results are summarized and com 
pared with those of Schmid and Wasser- 
man 1 and Nix and Schmid 2 in Table i. 

There is a difference between the present 
results on zinc and cadmium and those of 
the earlier investigators in that they did not 
observe the preference for a definite growth 
direction, presumably because too few 
(from 6 to 10, according to their report) 
crystals participated in producing the 
patterns. Orientation textures at and near 
the surface were not determined. 



The difference between the results on 
magnesium cannot be accounted for so 
simply. Possibly the texture of magnesium 
varies with the casting conditions; e.g., the 



Reference 
edge of 
spec/ men_ 




FIG. 17. STEREOGRAPHIC PROJECTION or 

ORIENTATIONS OF INDIVIDUAL GRAINS DESIG 
NATED BY CORRESPONDING NUMBERS IN FlG. 1 6. 

Poles of (100) planes are shown by open 
circles, poles of (ooi) planes by filled circles. 
Specimen surface is parallel to original surface 
of casting and is the plane of projection. 

rate of cooling. Scheil 5 reported such a 
difference for salts in discussing Nix and 
Schmid s paper. In the sample used here 
the rate of freezing must have been very 

TABLE i. Orientation Textures in Cast 

Poly crystalline Zinc, Cadmium and 

Magnesium 







Plane Parallel to 






Cooling Surface 


Metal 


Part of Casting 


Present 
Deter 


Reported by 
Schrnid and 
Wasserman 






mina 


and Nix and 






tion 


Schniid a 


Zinc 


Chilled surface 


(OOI ) 


Not stated 




Columnar grains 


(ioo) & 


(ooi) perpen 








dicular 




Intermediate 


(lQl)C 


None stated 




Air cooled sur 


(ooi) 


Not stated 




face 






Cadmium . . 


Chilled surface 


(ooi) 


Not stated 




Columnar grains 


(100)* 


(ooi) perpen 
dicular 


Magnesium. 


Columnar grains 


(205 )* 


(no)* 



* Furnished by the Dow Chemical Co. through the 

courtesy of Dr. J. D. Hanawalt. 



a The indices have been changed to conform to the 
system used here. 

6 Approximately, see later discussion. 

c Not always found. Where it has been observed it 
has been mixed with orientations described above or 
randomly oriented grains. 

d The text and table in Nix and Schmid s paper do 
not agree; their table gives this as (roo). 



GERALD EDMUNDS 



IQI 



slow; no experiments were conducted to 
investigate effects of other cooling rates. 
The orientation texture of magnesium is 
discussed further in the next section. 



dodecahedron are also along the cube axes, 
and completely analogous results follow, 
again demonstrated by experiment. 1 2 
Crystals of lower symmetry cannot be 




- Original top surface: (ooi) parallel to surface with wide but 
undetermined variation from this position (100) absent. 

*- About i% in. above bottom surface: (100) parallel to surface 
with wide but undetermined variation from this position. 



^-About l ViQin. above bottom surface: (100) i6and (ooi) 
16, about equal intensity.* 



-About Jfe in. above bottom surface: (100) 20 and (ooi) 
20, about equal intensity.* 

G4 jg^ VERTICAL SECTION THROUGH i Y\ C-INCH DEEP CASTING No. 3. 
Etch: concentrated HCL About natural size. Orientations at various levels are indicated at 
right. 

*This surface contained many twins, presumably developed during its preparation. The (ooi) 
reflections may have been due to them. See also explanatory note under Figs. 4 to 15, 

treated so simply. In the hexagonal close- 
packed metals, such, for example, as zinc, 
cadmium and magnesium, the planes of 
closest packing are the basal (ooi) planes, 
and these alone cannot completely bound a 
crystal. The next most closely packed 
planes are the second-order prism {no} 
planes. Based on consideration of these two 
types of planes alone, the Bravais rule, 
with some extension, would indicate that a 
hexagonal prism bounded by (ooi) and 
{no) planes would be the natural form; 
the ratio of the height to the thickness of 
the natural crystal would depend upon the 
relative atomic density on these two types 
of planes, the crystal being lower the 
greater the density on (ooi) planes relative 
to {no} planes. Since the ratio of these 
atomic densities is directly proportional to 
the axial ratio (c/a) of the crystallographic 
axes, the natural form of a crystal of large 
axial ratio would tend to be squatty, and 
one of small axial ratio tall. The six equal 
and longest dimensions, corresponding to 
the directions of maximum growth rate, 
through such a prism would be directions 



RATIONALIZATION OF THE ORIENTATION 
TEXTURES OF CAST METALS 

A. Bravais 6 has concluded that the faces 
most likely to bound a natural crystal are, 
in general, those of greater reticular 
(atomic) density. C. H. Desch 7 stated, on 
this basis, that "The more open the pack 
ing of a plane ... the greater its velocity 
of growth in a direction normal to itself, 
and the less its chance of survival during 
continued growth." 

The application of this rule to the cubic 
metals is simple. In face-centered cubic 
metals the octahedral {in} planes are 
those of closest packing. The longest dimen 
sions through an octahedron are along the 
directions of the cube axes, and hence by 
the Bravais rule the maximum growth rate 
would be along the [100] directions. The 
columnar grain axes of face-centered cubic 
metals have been found 1 2 to be approxi 
mately parallel to [iooj directions, in agree 
ment with the rule. In body-centered cubic 
metals the {110} planes are those of closest 
packing. The longest dimensions through a 



192 



GRAIN ORIENTATION OF CAST ZINC, CADMIUM AND MAGNESIUM 









^Original surface cast against preheated Transite: (ooi) and 
(100), very few reflections from either, 
in, in from surface cast against Transite: (100) 10 and 
some at large angles; (ooi), a few reflections. 






-i}4, in. in from surface cast against steel: (100) 14. 



-f| in. in from surface cast against steel: (100) 18. 



-0.201 in. in from surface cast against steel: (100) 23. 
-0.046 in. in from surface cast against steel: (ooi) + 7, also 
* v (100) and (101) reflections over entire angular range of test. 

{ .Original surface, cast against cooled steel: (ooi) 11. 

FIG. 19. HORIZONTAL SECTION THROUGH CASTING No. 4. 
Etch: concentrated HCL About 0.7 X. Orientations as determined on the indicated vertical 

sections are given at the right. 



GERALD EDMUNDS 



parallel to {no} planes, and, qualitatively, 
the angle between the hexagonal axis [ooi] 
and the direction of maximum growth rate 
would increase with an increase in the axial 




FIG. 20. SECTION THROUGH CASTING No. 5. 
Etch: concentrated HC1. X 6. The dark, 
rough rim is copper electroplate, applied to 
specimen to prevent loss of edge detail in 
polishing. Orientation textures at various 
levels are: 
DEPTH BELOW 
SURFACE, IN. ORIENTATION TEXTURE 

Original surface (ooi) tangent to surface 15 

to axis of bar 

0.009 (100) tangent to surface 25 

to axis of bar 

0.019 (100) tangent to surface 29 

to axis of bar 

> (100) tangent to surface 17 

to axis of bar 

ratio (c/d). It would seem merely to be 
coincidence if the direction of maximum 
growth rate happened to be either parallel to a 

TABLE 2. Experimental Relationships 



Metal 


Plane 
Approxi 
mately 
Normal 
to 
Columnar 
Grain Axis 


Approxi 
mate 
Angle of 
Growth 
Direction 
to [ooi] 


Axial Ratio 
(c/a) 


Cadmium .... 
Zinc 


Cioo) 
(ioo)* 


90 
90 a 


1.89 at 300C. 
i 886 at 400C 


Magnesium. .. 


(205) 


34 6 


1.623 at 25C. C 



None of the 20 grains whose orientation was 
determined individually (Fig. 17) had its plane in 
clined at less than 3, and only two had it inclined at 
less than 5 to the specimen surface. The number of 
determinations was too small to allow any particular 
significance to be attached to this observation. 

*> (ooi) A {205} = 36 52 . 

6 Axial ratio is changed but little with temperature, 
at least up to 200C. (ref. 8). 



principal cry stallo graphic direction or per 
pendicular to a principal crystallographic 
plane. 
Since the ratio of atomic density of (ooi) 




FlG. 21. A HORIZONTAL SECTION (UPPER 

PHOTOGRAPH), i.i INCH ABOVE THE BOTTOM 

SURFACE, AND A VERTICAL (LOWER PHOTO 
GRAPH) SECTION OF CADMIUM CASTING No. 6. 

Etch: concentrated HC1. Original magnifi 
cation 2; reduced approximately J^ in repro 
duction. 

Level a, i.i in. above bottom surface 
(too) 22. 

Level 6, original bottom surf ace (ooi) 17, 
strong; also random orientations. 

At the original top surface (not shown here) 
the (100) planes occupied a mean position 5 
from parallelism with the surface. The ob 
served range of orientations was 5 from this 
position so that the X-ray pattern showed two 
diffraction arcs from (100) that touched at the 
middle of the pattern. 

to { 1 10} is lower in magnesium than in zinc 
and cadmium, the above hypothesis led to 
the prediction, later confirmed by the deter 
mination reported on page 190, that the 
axis of the columnar grains of magne- 



GRAIN ORIENTATION OF CAST ZINC, CADMIUM AND MAGNESIUM 



slum would be parallel to a { noj plane and 
at an angle to the hexagonal axis smaller 
than the same angle for zinc. The experi 
mental relationships are shown in Table 2. 



about 1.6; according to the Bravais rule, as 
here extended, the orientation of columnar 
grains in castings of them is probably more 
like that of magnesium than of zinc and 




- - -"* , i 11 ! mumammmim*. jf M -HWJ 

FIG. 22 HORIZONTAL (UPPER) AND VERTICAL (LOWER) SECTIONS OF MAGNESIUM CASTING No 7 
Etch: horizontal section, 1:2 acetic acid and H 2 0; vertical section, 10 per cent ammonium 
persulphate m^HsO. Slightly reduced from natural size. Numbers designate grains whose indi 
vidual orientations are given by Fig. 23. 



No consideration has been given in 
the analysis to the effect of planes of 
slightly lesser atomic density. The only 
effect that would be anticipated from in 
cluding them would be to further diminish 
the predicted angle between the hexagonal 
axis [ooi] and the direction of growth, 
particularly in crystals of low axial ratio. 

Hexagonal close-packed metals, except 
zinc and cadmium, have an axial ratio of 



cadmium. A similar analysis would follow 
directly for other crystal systems. 

Attempts to rationalize the (ooi) surface 
orientation found for zinc and cadmium 
and the {101} intermediate orientation 
found for zinc have not been successful. 

CONCLUSIONS 

The interior of zinc castings consists of 
columnar grains aligned approximately 



GERALD EDMUNDS 



195 



parallel to the direction of the thermal 
gradient. The mean crystallographic orien 
tation of the columnar grains is such that a 
{iool prism "plane is approximately per 
pendicular to the long axis. Whether it is 
precisely perpendicular to the long axis 
cannot be determined from the available 
data; the hypothesis presented suggests 
that it is not. 

The mold surface of a zinc casting usually 
consists of relatively small grains with their 
(ooi) or basal planes approximately parallel 
to the mold surface. This is contrary to the 
assumptions made by other workers in this 
field and to the best of the writer s knowl 
edge is the first observation of this kind on 
cast metals. 

Intermediate between the surface and 
the long columnar grains frequently there is 
a region in which grains having a {101} 
pyramid plane perpendicular to the thermal 
gradient predominate. Wherever observed, 
this orientation has always been accom 
panied by the orientations described above 
or by randomly oriented grains. Occasion 
ally this orientation has been found at the 
mold surface. 

In cast cadmium the surface texture and 
the orientation of the columnar grains are 
as they are in zinc. The intermediate 
orientation found in zinc was not looked for 
in cadmium. 

In cast magnesium a {205} pyramid 
plane is approximately perpendicular to the 
long axes of the columnar grains. This dis 
agrees with results reported by Nix and 
Schmid. 

A hypothesis is developed rationalizing 
the orientation texture of the columnar 
grains. This hypothesis led to a correct 
prediction of the orientation texture of 
magnesium. It also predicts that most 
hexagonal close-packed metals will have an 
orientation texture more like magnesium 
than like zinc and cadmium. 

A new method of determining orienta 
tions of coarse-grained castings, designated 
as the rotating-translating specimen X-ray 



surface-reflection method, has been em 
ployed. With its use diffraction occurs from 
every grain in the specimen surface, greatly 
facilitating fiber-axis determinations on 




FlG. 23. STEREOGRAPmC PROJECTION OF 
ORIENTATIONS OF INDIVIDUAL MAGNESIUM 
GRAINS DESIGNATED BY CORRESPONDING NUM 
BERS IN FlG. 22. 

Poles of (205) planes have an average posi 
tion close (9) to the plane of projection and are 
shown by solid circles. Poles of (ooi) planes are 
shown by crosses and of (roo) planes by open 
circles. The specimen surface is perpendicular 
to the axis of the billet and is the plane of 
projection. 

coarse-grained specimens. The validity of 
the method has been confirmed by orienta 
tion determinations made by the back- 
reflection Laue method. 

ACKNOWLEDGMENTS 

The writer is indebted to his colleagues 
for assistance and counsel, and especially to 
Mr. M. L. Fuller for the X-ray examina 
tions and to Mr. H. S. Gruber for specimen 
preparation; to Dr. C. H. Mathewson for 
having incited interest in the problem, and 
to The New Jersey Zinc Co. for permission 
to conduct the investigation and publish 
the results. 

REFERENCES 

1. E. Schmid: Ztsch. Metallkunde (1928) 20, 373. 

2. F. C. Nix and E. Schmid: Ztsch. Metallkunde 



(1929) 21, 291. 
K. Weissenberg: Ztsch. Physik I 



3. K." Weissenberg: Ztsch. Physik (1921) 8, 20. 

4. C. H. Desch: Metallography, Ed. 4, 187. New 

York, 1937. Longmans Green & Co. 

5. E. Scheil: Ztsch. Metallkunde (1929) 21, 292. 

6. A. Bravais: Etudes crystallographiqu.es, 168. 

Paris, 1866. 

7. C. H. Desch: The Chemistry of Solids, 33- Cornell 



Univ. Press, 1934- 
8. E. Schmid: Ztsch. Ele, 



Ctrochem. (1931) 37, 447- 



GRAIN ORIENTATION OP CAST ZINC, CADMIUM AND MAGNESIUM 



DISCUSSION 

(L. W. Eastwood presiding) 

J. C. MCDONALD,* Midland, Mich. Some 
preliminary work has been done on magnesium, 
in an effort to check the author s finding. The 
result is concordant with his to this extent: that 
grains have been observed whose growth axes 
were not parallel to the (ooi) plane. 

The particular casting examined by the 
author was freezing at a rate of about i in. per 
min. at the point examined. The grains started 
to grow at a lower point in the casting, where 
the rate was higher by an unknown amount, 
probably not large. The uniformity of orienta 
tion observed in the surface, and its departure 
from the normal type explained by the author s 
theory, suggest a dependence of orientation on 
rate of cooling. At the surface of a liquid metal, 
the temperature gradient becomes extremely 
high, probably orders of magnitude higher 
than the gradient present in the interior of 
even the rapidly cooling die casting. This effect 
would exist even at an air-liquid interface. 
Under such conditions, we may speculate 
that the superficial layer of grains is forced to 
form in an orientation different from that 
normally occurring at lower rates of freezing. 
What is the author s reaction to this suggestion? 



A. B. GRENiNGERjt Cambridge, Mass. Has 
Mr. Edmunds made observations on the 
dendritic habits of the hexagonal metals he 
has studied? I believe that dendritic growth 
and preferred orientation in castings must be 
intimately allied and a rationalization of the 
latter that will be worth while cannot be made 
until there are available a few positive state 
ments regarding dendritic habit. In other 
words, the rationalization of orientation 
textures of cast metals should preferably be 
made first with reference to dendritic growth 
rather than to geometry of the crystal. 

One of the few statements regarding habit 
that mineralogists accept as reasonably accu 
rate is: Habit is a resultant of (i) the geometry 
of the species, and (2) its environment during 
the process of crystallization. The generaliza 
tion of Bravais, and its extension by Mr. 
Edmunds, considers only geometry; hence it 



is not surprising that the application of the 
rule does not yield wholly satisfactory results. 
In natural isometric metal crystals, the cube 
is even more common than the octahedron, and 
thus, according to the author s reasoning, the 
direction of most rapid growth should be 
<iu>. Actually, in natural isometric metal 
crystals, both <ni> and <ioo> are com 
mon dendrite axes, but for cubic metals and 
alloys prepared in the laboratory, the dendrite 
axes are apparently always <ioo>. There is 
little or no published information on this point; 
measurements have been made at Harvard on 
several primary solutions of copper and of iron 
and no axes other than <ioo> have been 
found. 

It is to be expected, therefore, that columnar 
crystals of cubic metals should be oriented with 
the long axes nearly parallel to <ioo>, for 
the grain orientations that survive during 
growth must be those whose natural direction 
of maximum growth velocity coincides with 
the direction of the temperature gradient. I 
know of no published data on the dendritic 
habits of the hexagonal metals. Judging from 
Mr. Edmunds results on orientation texture, 
zinc and cadmium should have <2io> parallel 
to the dendrite axes and magnesium would have 
a considerably different dendritic habit. 

C. H. MATHEWSON,* New Haven, Conn. 
With reference to Dr. Greninger s remarks 
concerning the crystalline habit of metal 
crystals, it may be of interest to note that silver 
crystallizes from dilute solution in sodium 
in the form of w T ell developed simple octahedra. 
Primary silver crystals obtained from a mixture 
containing about 75 per cent of silver bore 
great resemblance to specimens of native silver, 
occurring, according to Dana, often in 
groups branching at 60 parallel to the diagonals 
of an octahedral face." Photographs and 
sketches of these silver crystals are given in my 
paper on sodium-silver alloys. 9 

G. EDMUNDS (author s reply). Mr. McDon 
ald has speculated upon the origin of the surface 
texture of zinc and cadmium, w r hich was found 
to differ from that of the interior of the castings, 
and has raised the question as to whether this 



* The Dow Chemical Co. 
t Assistant Professor of 
University. 



Metallurgy, Harvard 



* Professor of Metallurgy, Yale University. 
9 C. H. Mathewson: Int. Ztsch. Metallographie 
1911) i, 51-63. 



DISCUSSION 



197 



might be due to a tendency of a different 
orientation texture to develop under an ex 
treme temperature gradient. This possibility 
was considered, but all experiments conducted 
in attempts to gain support of this idea gave 
negative results. It was found that the depth 
to which this orientation persisted did not 
increase under conditions of rapid cooling, as 
by die casting. It was also found that the 
surface of a casting made in an open tray and 
cooled in still air had this surface texture. In 
the latter case the metal was somewhat super 
heated for pouring; under such conditions 
there would be no instantaneous metal crystal 
lization at the surface. It is doubted that an 
extreme temperature gradient existed in the 
surface layer of the relatively high thermal 
conductivity melt. It seems unlikely, therefore, 
that the difference between surface and interior 
texture can be attributed to differences in 
thermal gradient. Several other hypotheses 
have also been considered in attempts to find 
one that would rationalize the surface-orienta 
tion texture of zinc and cadmium, but none 
has been successful. 

I am in complete agreement with Dr. 
Greninger s statement that dendritic growth 
and preferred orientations in castings are 
intimately allied; in fact, I would go further 
and say that both are manifestations of the 
same influence that is, the tendency for 
crystals to grow in certain directions more than 
in others. The hypothesis in the paper ration 
alizes this tendency from purely geometrical 
concepts. The other important influence is the 
interference of growing dendrites upon one 
another, allowing survival only of those having 



a direction of rapid growth nearly parallel to 
the thermal gradient. 

As far as observations of dendrite habits for 
the hexagonal metals are concerned, these have 
been made on both zinc and cadmium. Rather 
indistinct dendrites are to be seen on Figs. 3, 
1 6 and 18. The general pattern is of noniso- 
morphous impurities having been deposited in 
films parallel to basal planes. Unrestricted 
growth yields dendrites, whose trunks and 
branches have as their axis (at least approxi 
mately) a direction [210]. 

In his other comment Dr. Greninger states 
that "in natural isometric metal crystals, the 
cube is even more common than the octa 
hedron" and that thus, according to the 
hypothesis in the paper the direction of most 
rapid growth should be [in]. This statement 
is contrary to that of Guillet and Portevin; 10 in 
reporting upon the classic work of Tschernoff 
and of Osmond, they state that "with pure 
metals and solid solutions the development 
nearly always occurs along the axes of an 
octahedron* forming tree-like structures known 
as crystallites or dendrites . . . which, if 
allowed to develop freely and at the same speed 
in all directions, would finally take the form 
of an octahedron." 

Dr. Greninger later goes on to say that "for 
cubic alloys prepared in the laboratory, the 
dendrite axes apparently are always [100]." 
This, of course, is in complete agreement with 
the older results referred to and the hypothesis 
that has been presented. 

10 Guillet and Portevin: Metallography and 
Macrography, 69. London, 1922. G. Bell & Sons, Ltd. 

* "Most metals crystallise in the cubic system, the 
directions of the dendrite branches being the qua 
ternary axes of the cube." 



Studies upon the Corrosion of Tin Effects of Cations in 
Carbonate Solutions and Effects of Alloying Elements 

BY GERHARD DERGE,* MEMBER A.I.M.E., AND HAROLD MARKUS* 



(New York Meeting, February 1941) 



THE first paper 1 of this series described a 
technique of careful surface preparation by 
means of which reproducible results may be 
obtained from potential measurements of 
the behavior of tin in carbonate solutions 
with a pH range from 8.4 to 11.2. It then 
went on to establish the behavior of pure 
tin electrodes under these conditions and 
showed that the metal remains active at all 
times, though the steady potential attained 
after several hours decreases from i.iaf 
volts at a pH of 12.6 to 0.6 volts at a pH of 
8.4. This behavior may serve as a norm 
with which all other data can be compared. 
In the second paper, 2 small amounts of 
other anions were added to the carbonate 
solutions and the conclusion was reached 
that the common impurities encountered in 
alkaline materials packaged in tin do not 
increase corrosion. In fact, in some cases 
these anions may actually promote the 
formation of protective films. 

The present paper rounds out this survey 
by presenting data on the influence of 
cations added to the carbonate solutions, as 
well as the effect of alloy additions to the 
tin. In many cases the potential data are 
supplemented with weight-loss measure 
ments. Information is also presented on the 

Manuscript received at the office of the Institute 
Nov. 30, 1940. Issued as T.P. 1306 in METALS 
TECHNOLOGY, April 1941. 

* Metals Research Laboratory, Carnegie Institute 
of Technology, Pittsburgh, Pa. 

1 References are at the end of the paper. 

f Throughout this series potential values are 
referred to a saturated calomel cell as zero, 
with noble metals like silver negative; i.e., 
increasing positive values indicate a greater 
tendency to corrode. 



effects of some additional anions, the influ 
ence of gases above the electrolyte, the 
behavior of high-purity tin, and the effect 
of atmospheric oxidation on the corrosion 
resistance of the above-named alloys. 
Analysis of the entire body of data discloses 
some general rules applicable to corrosion in 
carbonate solutions. The experimental 
methods used are the same as those previ 
ously described. 

EXPERIMENTAL DATA 

Cations. The carbonates of the cations 
K + , Li + , NH 4 "*", which are soluble, were 
studied in the same way as the anions, by 
adding o.oi mol of the metal carbonate to 
one liter of o.i M sodium carbonate solu 
tion, pH 1 1. 2, or to o.i M sodium bicarbon 
ate, pH 8.4. The other metallic carbonates 
are only slightly soluble, of the order of io~ 3 
mols per liter or less, and the sodium car 
bonate solutions were saturated by adding 
excess reagent several days before the ex 
periment, shaking frequently, and filtering 
just before use. The chemicals were c.p. 
grade in each case except the indium salt. 
This salt was prepared by solution of the 
metal in hydrochloric acid and precipitation 
by addition of Na 2 C0 3 . These cations may 
be divided into two classes, according to 
their effect on the potential behavior of the 
tin: (i) normal, (2) passive. 

To offer a basis of comparison, curves i of 
Figs, i and 2 are given to illustrate the 
behavior of tin in carbonate solutions 
saturated with CaC0 3 . These curves are the 



198 



GERHARD DERGE AND HAROLD MARKUS 



IQ9 



same as those obtained from solutions 
without Ca ++ additions and are classed as 
normal. The same curves also typify the 
effect of additions of In +++ , Li + , Fe ++ , 



anions that were not reported in the previ 
ous paper will be described here. Arsenate 
ion falls in class i above and in the previous 
paper; 2 i.e., it has little or no effect upon 



0.8 



0.6 



0.2 - 




180 220 



TIME IN MINUTES 

JT IG !. TIME-POTENTIAL CURVES SHOWING 
INFLUENCE OF ADDED CATIONS ON BEHAVIOR OF 

TIN IN CARBONATE SOLUTIONS WITH PH II. 2. 

Curve i, Ca++ Curve 5, Pb^ 

Curve 2, Mg^ Curve 6, Ni++ 

Curve 3, Cu^ Curve 7, Ag + 
Curve 4, Zn"*" 1 " 

Cd ++ , Mn ++ , Ba ++ , and Bi +++ at both pH 
values, and the behavior of K + and Co ++ at 
a pH of 1 1. 2 only. 

The cations which promote passivity at 
both pH values are Mg ++ , Cu+ + , Zn ++ , 
Pb ++ , Ni ++ , and Ag+. The influence of Mg ++ 
and Ni ++ at a pH of 11.2 is very small, and 
it is hardly correct to say that the ions 
induce passivity. However, they do change 
the shape of the normal curve in this direc 
tion. At the lower pH, K + and Co+ + are also 
passivating. Time-potential curves illus 
trating these effects are shown in Figs, i 
and 2. 

Anions. In order to present as complete 
a picture as possible, the effects of a few 



0.6 



0.4 



0.2 




(ro&soo) 



20 60 100 140 180 220 

TIME IN MINUTES 

FIG. 2. TIME-POTENTIAL CURVES SHOWING 

INFLUENCE OF ADDED CATIONS ON BEHAVIOR OF 
TIN IN CARBONATE SOLUTIONS WITH PH 8.4. 

Curve i, Ca** Curve 6, Ni++ 

Curve 2, Mg ++ Curve 7, Ag + 

Curve 3, Cu ++ Curve 8, K + 

Curve 4, Zn** Curve 9, Co ++ 
Curve 5, Pb++ 

the normal curve. Arsenite ion is in class 2 
of the previous paper; 2 it induces passivity 
at a pH below 10.0. Silicate is in class 3 in 
that it promotes passivity over the entire 
pH range studied. Sulphide ion requires a 
new classification in that it promotes 
activity throughout the pH range studied. 
The important time-potential curves for 
these ions are shown in Fig. 3. 

Influence of Atmosphere. A brief de 
scription of the influence of the atmosphere 
above the solution gives information on the 
mechanism of corrosion. To demonstrate 
this, an experiment was made in which the 
gas passed through the cell could be alter- 



200 



STUDIES UPON THE CORROSION OF TIN 



nated between tank oxygen and tank 
nitrogen. The time-potential curve in a 
solution with pH 11.2 is shown in Fig. 4. 
While nitrogen is being used, the metal is 




0.2 - 



60 



140 180 220 



TIME IN MINUTES 
FIG. 3. TIME-POTENTIAL CURVES SHOWING 

INFLUENCE OF ADDED ANIONS ON BEHAVIOR OF 

TIN IN CARBONATE SOLUTIONS. 

Curve i, o.oi M Na^HAsOs, pH 10, 
Curve 2, i.o per cent Na2SiOs; pH 11.2. 

( 4 o-42 Be\ Tech.) 

Curve 3, o.i per cent NasSiOg, pH 11.2. 
Curve 4, o.i per cent Na2SiOs, pH 10.0. 
Curve 5, o.oor M Na 2 S, pH 11.2. 

abnormally active, with a steady potential 
of 0.98 volts. In changing to oxygen the gas 
was bubbled through the electrolyte for 
% hr. and then maintained over the elec 
trolyte in the normal manner. The potential 
falls off at once, and a steady value of 
0.1780 volts was shown for a period of 3 
days. When nitrogen was again admitted 
the potential rose rapidly to its previous 
value. Evidently the excess of oxygen 
stabilizes the stannic hydroxide corrosion 
product to the extent that it becomes 
protective. The total absence of oxygen 
caused_^by the introduction of nitrogen 



makes the oxide film nonprotective. These 
effects set up a differential aeration cell in 
ordinary weight-loss experiments in which 
the lower part of the specimen, away from 



0.9 



0.7 



0.5 



START OF 0, 




120 



200 280 360 



TJME IH MINUTES 

FIG. 4, EFFECT OF OXYGEN AND NITROGEN 
ON THE TIME -POTENTIAL CURVE FOR CHEMPUR 
TIN IN CARBONATE SOLUTIONS, PH 1 1. 2. 

^The system was under an atmosphere of 
nitrogen for the first 220 minutes; the atmos 
phere was then changed to oxygen. 

atmospheric oxygen, becomes anodic to the 
upper part, where a protective film is 
formed. 

Preparation and, Selection of Alloys. In 
examining the effects of alloying elements, a 
number of factors were considered in 
preparing the samples. The choice of alloy 
ing elements was not limited to those which 
might have commercial possibilities, since 
the principal object was to obtain a wide 
variety of data upon which to base general 
conclusions. The problem was confined to 
finding the effects of small amounts of 
alloying elements on the corrosion of tin, 
and examination of entire alloy systems did 
not seem pertinent. The compositions were 



GERHARD DERGE AND HAROLD MARKUS 



2OI 



selected with the idea of obtaining a typical 
variety of structures from a given alloy 
system. For example, a hypoeutectic, a 
eutectic, and a hypereutectic composition 



lies so close to pure tin that the selection 
desired could not be made complete with 
any certainty. In other cases the eutectic 
composition was so far from the tin-rich 



TABLE i. Compositions 



Alloying Element, Per Cent 


Alloy Type 


Source of Element 


Preparation 




Hypoeutectic 


Chempur Sn 


Melted in clay crucible under char 




Eutectic 




coal 19.71 per cent Cu master 




Hypereutectic 


O F H C Cu 


alloy 




Hypereutectic 


99.9+ % Cu 




5 o Cu 


Hypereutectic 
















Solid solution 




/Melted in clay crucible under char- 


o e Zn 


Hypoeutectic 


Chempur Sn 


1 coal 24.37 per cent Zn master 




Hypoeutectic 


Horse Head Special Zn 


) alloy 


g 2 Zn 


Hypoeutectic 








Hyp ereutectic 


99-95 % Zn 


) 


15 o Zn 


Hypereutectic 




^ 










o.i Ni 


Hypoeutectic 


/Chempur Sn 


/ Melted in clay crucible under char- 


0.18 Ni 


Eutectic 


< 


) coal 10.0 per cent master alloy 




Hypereutectic 


) Electrolytic Ni 


| 






V99-95 % Ni 


V 


o.i Fe 


Hypereutectic 


/Chempur Sn 


{Melted in clay crucible under char 
coal 2.6 per cent master alloy 


T o Fe 


Hypereutectic 


\ HrLz purified Carbonyl 








iFe 99-95 % Fe 




o.i Pb 


Solid solution 


/Chempur Sn and 


/Melted in clay crucible under char 


I 5 Pb 


Solid solution 




coal 24 98 per cent master alloy 


3 .o Pb 


Hypoeutectic 


} High Purity Pb 




5 o Pb . 


Hypereutectic 


\99999 % Pb 


\ 










0.05 Ag 


Solid solution 


/Chempur Sn and 


/Melted clay crucible under char 


o i Ag 


Hypoeutectic 


} 


coal 5.60 per cent Ag master alloy 




Hypoeutectic 


\ High Purity Ag 


) 






199.995 % Ag 


V 


o i In 


Solid solution 


( Chempur Sn and 


( Melted in clay crucible under char- 


i o In 


Hypoeutectic 




^ coal 3 per cent master alloy 


3 o In . 


Hypoeutectic 


( 99 986 % In 










^ 


o i Mg 


Hypoeutectic 


/Chempur Sn and 


/Melted in clay crucible under flux 


i 87 Mg 


Eutectic 




) CaFi, 40 grams; KC1, 320 grams; 


2 5 Mg 


Hypereutectic 


"\ High Purity Mg 


) NaF, 60 grams; MgCh, 900 grams; 






1 99.9 % Mg 


V. 10 per cent master alloy 


0.5 Cd 


Solid solution 


( Chempur Sn and 


( Melted in glass tubing under 


i o Cd 


Solid solution 




< paraffin 2 5 per cent master alloy 


2 5 Cd 


Hypoeutectoid 


( 99-9 % Cd 










^ 


o 5 Sb . ... 


Solid solution 


/Chempur Sn and 


/ Melted in glass tubing under 


i o Sb 


Solid solution 




j paraffin 6 per cent master alloy 


3.0 Sb 


Solid solution 


) 99.60 % Sb 


\ 


6 o Sb 






( 




\ ~ pt 


^ 




o.i Bi. . 


Hypoeutectic 


/Chempur Sn and 


/ Melted in clay crucible under char- 
} coal i per cent master alloy 


o c Bi 


Hypoeutectic 


\ Pharmaceutical Bi 


j Melted in glass tubing under 


1,0 Bi ... 


Hypoeutectic 


V99-98 % Bi 


v, paraffin i per cent master alloy 










o.i Ca 


Hypereutectic 


( Chempur Sn and 


( Melted in plumbago crucible under 
< charcoal i per cent master alloy 






( Commercial Pure Ca 




High-purity Sn 




Special preparation 


f Melted inAUOs crucible in vacuum. 








( Estimated 99.9999 per cent pure 



were selected whenever possible. In some 
cases compositions within and without the 
terminal solid solution range were selected. 
In many systems the eutectic composition 



side that an examination of other than hypo 
compositions was omitted. It is believed, 
however, that a sufficient variety has been 
studied to disclose any important generali- 



2O2 



STUDIES UPON THE CORROSION OF TIN 



ties that may exist. The compositions used 
are listed in Table i. Most of the alloys 
were melted in clay crucibles under char 
coal. Master alloys were prepared first, and 



BEHAVIOR OF ALLOY 

Potential measurements were made at 
the two extremes of the pH range only 



0,8 



0.6 







0.4 



0.2 - 




180 220 



Tl ME IN MINUTES 

FIG. $. TIME-POTENTIAL CURVES FOR NORMAL 
ALLOYS IN CARBONATE SOLUTIONS. 
Curve i, 1.5 per cent Pb, pH 84. 
Curve 2, 1.5 per cent Pb, pH 11.2. 
Curve 3, o.i per cent Fe, pH 8,4. 
Curve 4, i per cent Sb, pH 8.4. 
Curve 5, 6 per cent Sb, pH 9.5. 
Curve 6, 0.5 per cent Cd, pH 8.4. 
Curve 7, o.i per cent Ni, pH 8.4. 

the final samples were made from them. 
The composition was taken as that of the 
weighed elements, since exact values were 
not important. The structures were checked 
with those to be expected from the equi 
librium diagram. In some special cases 
chemical analyses were made. Important 
exceptions to these generalities are listed in 
Table i. The surfaces of the potential 
specimens were finally prepared as de 
scribed previously, 2 except that the final 
vacuum anneal was always 10 below the 
eutectic temperature. 



o. s 



0.6 



Uj 



0.2 




20 



60 



100 



180 220 



TIME IN MINUTES 

FIG. 6. TIME-POTENTIAL CURVES FOR TIN- 
COPPER ALLOYS IN CARBONATE SOLUTIONS. 

Curve i, 0.75 per cent Cu, pH 11.2. 
Curve 2, 0.75 per cent Cu, pH 10.0. 
Curve 3, 0.75 per cent Cu, pH 9.5. 
Curve 4, 0.75 per cent Cu, pH 8.4. 
Curve 5, 3 per cent Cu, pH 10.0. 
Curve 6, 3 per cent Cu, pH 9.5. 
Curve 7, 0.4 per cent Cu, pH 10. 

(pH 1 1. 2 and 8.4) with intermediate values 
for a few cases of special interest. The 
results may be divided into three groups: 
(i) normal, (2) active, and (3) passive, and 
will be described in that order. 

To establish the validity of using the 
curves obtained from a good grade of 
commercial tin as a standard of comparison, 
a series of measurements was made with a 
special grade of high-purity tin prepared in 
the laboratory for extrusion work. 3 The 
material has been estimated as six nines 
pure. The potential curves for this do not 



GERHARD DERGE AND HAROLD MARKUS 



203 



differ in any significant features from those 
already published for Chempur tin, al 
though it seemed as though the final, 
steady, horizontal section of the curve is 




10 



40 



20 30 

TIME IN DAYS 

FIG. 7. WEIGHT-LOSS OF TIN-COPPER ALLOYS 

AT pH 1 1. 2 AND 10.0. 
Curve i, 5 per cent Cu alloy, pH 11.2. 
Curve 2, 3 per cent Cu alloy, pH 11.2. 
Curve 3, 0.75 per cent Cu alloy, pH 11.2. 
Curve 4, i.o per cent Cu alloy, pH 10.0. 
Curve 5, 0.75 per cent Cu alloy, pH 10.0. 

reached more rapidly with the high-purity 
material. 

i. Normal. The alloying elements that 
do not change these normal curves mark 
edly are Cd, Fe, Sb, Cu, Ag, Ni, Bi, and Pb. 
These include the commonest impurities in 
commercial grades of tin and substantiate 
the conclusion of the previous paragraph 
that small amounts of these impurities do 
not alter the corrosion characteristics of the 
material appreciably. It is necessary to add 
that Ag and Pb are the only ones of the 
elements listed that are entirely normal. At 
a pH of 8.4 the alloys with Sb, Cu, Cd, o.i 
and o.i 8 per cent Ni, and o.i per cent Fe 
are passive, and with the 6 per cent Sb alloy 



this effect persists up to a pH of 9.5. These 
deviations from normality are shown in 
Fig. 5. The alloys have been classed as 
normal, to avoid confusing the general 



feoo - 



-800- 



3 

k. 



400 




10 20 30 40 

TIME IN DAYS 

FIG. 8. WEIGHT-LOSS OP ANTIMONY ALLOYS IN 

CARBONATE SOLUTIONS. 
Curve i, 6 per cent Sb alloy, pH 11.2. 
Curve 2, 3 per cent Sb alloy, pH 11.2. 
Curve 3, i per cent Sb alloy, pH 11.2. 
Curve 4, 6 per cent Sb alloy, pH 10.0. 
Curve 5, 3 per cent Sb alloy, pH 10.0. 
Curve 6, i per cent Sb alloy, pH 10.0. 

picture, and because even pure tin is not 
attacked appreciably at so low a pH. 

The copper alloys are so widely used in 
practice that they were examined more 
completely. Some of the potential measure 
ments are shown in Fig. 6. There are no 
striking effects, but some differences occur, 
which, combined with other information to 
be described later, indicate that it is 
desirable to keep the copper content below 
the eutectic composition of 0.75 per cent 
Cu. The passive tendency shown by tin and 
most of its alloys at the low pH of 8.4 is 
extended to somewhat higher pH values by 
the addition of small amounts of copper. At 
a pH of 9.5 or 10 the potential curves for a 



2O4 



STUDIES UPON THE CORROSION OP TIN 



3 per cent Cu alloy correspond very closely 
to those of tin, but the alloys with 0.75 per 
cent or less of copper all have somewhat 
lower values in this range. When the pH is 




100 



220 



T/ME JN MINUTES 

FIG. g. TIME-POTENTIAL CURVES FOR TIN- 
CALCIUM AND TIN-MAGNESIUM ALLOYS IN 
CARBONATE SOLUTIONS. 

Curve i, 2.5 per cent Mg alloy, pH 11.2. 
Curve 2, 2.5 per cent Mg alloy, pH 84. 
Curve 3, o.i per cent Mg alloy, pH 8.4. 
Curve 4, o.i per cent Ca alloy, pH 11.2. 
Curve 5, o.i per cent Ca alloy, pH 8.4. 

as high as 11.2, all of the alloys show 
normal curves except 5 per cent Cu, which 
is somewhat higher. Weight-loss data at a 
pH of 1 1.2 show that the influence of copper 
content is negligible under these conditions. 
However, at a pH of 10.0 this effect be 
comes better defined (Fig. 7). It is evident 
in all of these diagrams that the influence of 
pH is much greater than that of alloy com 
position in this series. However, if the pH is 
predetermined it should be possible to 
minimize corrosion by proper choice of 
composition. 

Weight-loss curves have also been made 
for the antimony alloys (Fig, 8). As with 



copper, the amount of corrosion increases 
with the alloy content, as well as with the 
pH of the solution. Under the conditions of 
the test the antimony alloys are somewhat 



0.8 



0.6 



0.4 



0.2 




60 



100 140 130 220 



FIG. 10." 

ZINC AND 
SOLUTIONS, 

Curve i 
Curve 2 
Curve 3 
Curve 4 : 
Curve 5 
Curve 6 



TIME IN MINUTES 

. TIME-POTENTIAL CURVES POR TIN- 
TIN-INDIUM ALLOYS IN CARBONATE 



, o.i per cent Zn alloy, pH 11.2. 
, o.i per cent Zn alloy, pH 8.4. 

15.0 per cent Zn alloy, pH 11.2. 
,, o.i per cent In alloy, pH 11.2. 
, i.o per cent In alloy, pH 11.2. 
, o.i per cent In alloy, pH 8.4. 



superior to copper alloys. However, the 
corrosion product is a black, nonadherent 
material, which would be objectionable in 
most circumstances. 

2. Active. Of all the alloys studied, only 
those with the alialine earths, magnesium 
and calcium, were more severely corroded 
than pure tin. At a pH of 11.2 the steady 
potential values of both materials are about 
o.i volt higher than normal, but the initial 
values are very much higher, and it may be 
presumed that the low solubility of the car 
bonates affords some protection to the 
metal. At a pH of 8.4 the Ca alloy and the 



GERHARD DERGE AND HAROLD MARKUS 



205 



o.i per cent Mg are definitely passive, but 
the two magnesium-rich alloys have unusu 
ally high initial and steady values (Fig. 9). 
These magnesium-rich alloys liberate hy- 



i.o 



0.8 



$ 

k 



0.6 



0.4 




I 



20 



220 



60 100 I4O /SO 

Tl ME IN Ml NUTES 

FIG. ii. TIME-POTENTIAL CURVES FOR TIN- 
ZINC ALLOYS IN SODIUM HYDROXIDE SOLUTION 
PH II. 2. 

Curve i, 15 per cent Zn alloy. 
Curve 2, 10 per cent Zn alloy. 
Curve 3, 8 per cent Zn alloy. 
Curve 4, 2 per cent Zn alloy. 
Curve 5, 0.5 per cent Zn alloy. 

drogen spontaneously from the carbonate 
solutions and disintegrate rapidly, in fact, 
after a few weeks exposure to the atmos 
phere they become very brittle. The o.i per 
cent Mg alloy, which contains all the 
magnesium in solid solution, does not have 
this tendency to become brittle. This 
behavior of magnesium alloys is comparable 
to that of tin-rich aluminum alloys. 6 

3. Passive. The group of passive alloys 
includes zinc and indium, and they have 
been studied over the entire pH range. 
Since these alloys are important because of 
their ability to withstand corrosion in 
certain alkaline solutions, special attention 



has been given them. Typical potential 
curves are shown in Fig. 10. As little as 
0.025 pe r cent Zn or 0.5 per cent In is 
sufficient to cause passivity at a pH of 11.2. 



i.o 



0.8 



0.6 



0.4 




I 



I 



I 



20 60 100 J40 180 220 

T/ME TN MINUTES 

FIG. 12. TIME-POTENTIAL CURVES FOR TIN- 
ZINC ALLOYS IN SODIUM HYDROXIDE SOLUTION, 

pH 8.4. 

Curve 6, 8 per cent Zn alloy. 
Curve 7, 15 per cent Zn alloy. 
Curve 8, 10 per cent Zn alloy. 
Curve 9, 2 per cent Zn alloy. 
Curve 10, 0.5 per cent Zn alloy. 

The potential measurements have been 
checked by weight-loss experiments with i 
and 3 per cent In alloys at pH 10 and pH 
ii. 2, which were discontinued at the end of 
no days because no corrosion could be 
detected. Weight-loss experiments were 
made with the entire range of zinc alloys at 
a pH of 1 1. 2 and these were discontinued 
at the end of 55 days because no corrosion 
had occurred. In carbonate-free solutions 
of pH ii. 2 and 8.4 containing NaOH only, 
the series of zinc alloys remained passive 
in the solution of lower pH only (Figs, n 
and 12). This behavior of the zinc alloys 
indicates that a zinc carbonate film may be 



206 



STUDIES UPON THE CORROSION OF TIN 



responsible for inducing passivity in car 
bonate solutions, since passivity is absent in 
the NaOH solution of pH 11.2. In service 
tests, the Sn-Zn alloys withstood the 
corrosion of the alkaline products, but were 
subject to intergranular corrosion by the 
atmosphere. Neither the indium nor the 
zinc alloys appear to have commercial 
promise because the atmospheric corrosion 
of zinc is objectionable and indium must 
still be regarded as a precious metal. 

4. Oxide Films. The general importance 
of thin films in corrosion processes is well 
known. Most tin packages are baked at 
some time during their manufacture to 
facilitate drying of the lacquers used on the 
outside. Therefore, it seemed desirable to 
determine the effect of the alloys studied 
upon the atmospheric oxidation of the 
material, and also to determine the corro 
sion resistance of the oxide films thus 
formed. Two treatments were selected for 
this purpose: (i) baking for 50 min. at 
io5C., which approximates the treatment 
generally used in practice; (2) baking for 
5 hr. at i75C. The second treatment 
produces a thicker film, generally in the 
temper color range, which it was felt might 
be better suited to the type of observations 
to be described. None of these samples 
showed a corrosion behavior markedly 
different from that of the unoxidized metal, 
but in many cases the steady potential 
value was attained slowly. 

INTERPRETATION 

Classification of Alloys. The collection 
of corrosion data is relatively simple as 
compared to their correct interpretation. 
Nevertheless, an effort to derive as much 
information as possible from such data is 
usually in order. A general division of the 
alloys studied into good, bad, and indiffer 
ent groups may be made from the classifica 
tion into passive, active, and normal 
groups, which already has been made on the 
basis of the potential measurements. Zinc 
and indium may be considered as good over 



the entire pH range, magnesium and 
calcium are definitely bad, and the other 
metals are indifferent except that some of 
them do reduce corrosion at lower pH 
values. The weight-loss data are in general 
agreement with the potential measure 
ments and also show that antimony, even 
though it does not prevent corrosion, causes 
less corrosion than copper. Also, pure tin is 
more resistant to corrosion than the alloys 
with the bad or indifferent metals, but less 
resistant than its alloys with indium or 
zinc. 

Influence of Structure. The alloy com 
positions were selected so as to give as great 
a variety of structures as possible. The 
results do not point to any important 
general influence of this factor. Wherever 
the corrosion characteristics change with 
composition, they seem to change in pro 
portion to the amount of alloy element 
present, with no sharp changes as one 
passes from a hypoeutectic to a hyper- 
eutectic alloy. When solid solution occurs, 
the results are somewhat more surprising, 
and it may be that more accurate studies 
would reveal a sharp change when the solid 
solubility of the system is passed. The 
recent work of Brown, Fink, and Hunter 4 
bears out this point. The determinations of 
solid solubility at room temperature are 
somewhat uncertain and in view of the 
extremely limited range in most tin alloys, 
it may be that some of the alloys have not 
been classified properly. However, it seems 
quite certain that the alloys containing 0.5 
per cent, 5 per cent, i.o per cent, and 3.0 
per cent Sb are within the solid solution 
range and the changes in corrosion resis 
tance seem to be quite regular right on up 
to the two-phase 6 per cent Sb alloy. 

Periodic Relations. Checking the ele 
ments used as alloy additions upon a 
periodic table offers an opportunity to 
judge the influence of atomic structure 
upon corrosion of alloys. The series Ag, Cd, 
Sn, In, Sb have atomic numbers 47-51, and 
the new electrons are filling in the outer 



GERHARD DERGE AND HAROLD MARKUS 



207 



fifth quantum group in order. Additions of 
Ag have no influence on the potential 
behavior of Sn, but both Cd and Sb 
promote passivity at a low pH, while In 
induces passivity at a high pH also. In the 
series Fe, Ni, Cu, Zn the same general 
observations apply and as the structure of 
the atom becomes more like that of Sn, the 
corrosion resistance of the alloy increases. 
The alkaline earths Mg and Ca are far 
removed from Sn in the table and their 
alloys are readily attacked. The heavier 
metals Pb and Bi are very similar to Sn in 
structure and do not alter its corrosion 
properties greatly. Though this analysis 
does not explain why In and Zn promote 
passivity, it does indicate the desirability of 
alloying with chemically similar elements. 

Influence of Position in Electromotive 
Series. The position of the elements in the 
electromotive series is important and must 
be included in the interpretation of the 
results of the present work. The addition of 
alloying elements may change the electro 
chemical potentials of the alloy so as to 
affect the current flow between local ele 
ments. With tin, where very few elements 
alloy with it to form solid solutions, the 
general case finds the alloys of tin consisting 
of two constituents. If the second phase is 
less noble than the tin, local areas will differ 
electrochemically, setting up small cells and 
thus increasing the rate of corrosion. This 
also applies to phases that are nobler than 
tin. Where the alloying elements are elec 
trochemically similar, small cells may be 
formed but the currents are reduced to 
small values and the corrosion rate is not 
appreciably changed. 

The elements classed as normal in a 
previous section (Cd, Fe, Sb, Cu, Ag, Ni, 
Pb, and Bi) all lie reasonably close to Sn 
in the electromotive series except Ag. Al 
though each of these elements shows a 
potential curve similar to that of Sn, 
weight-loss measurements would perhaps 
reveal that all these elements increase the 



corrosion rate of Sn as shown for the Sn-Cu 
alloys. The elements classed as active (Mg 
and Ca) differ greatly from Sn electro 
chemically and, as expected, increase the 
corrosion rate tremendously. The passive 
class of elements (In and Zn) are situated 
near Sn in the series. However, passivity 
cannot be directly attributed to the position 
of these two elements in the series, but their 
nearness to Sn would be an advantage and 
would be one of the contributing factors in 
producing passivity. 

With respect to the cations in solution, 
something may be said about the influence 
of their position in the e.m.f. series. The 
cations of metals that are cathodic to tin 
should plate out on to the surface of the 
tin, rendering it passive if the coating 
possesses the proper characteristics. Like 
wise, the ions of metals that are anodic to 
tin will tend to remain in solution. Of the 
cations producing passivity, Cu ++ , Ag + , 
and Pb ++ are the most prominent and are 
cathodic to Sn. But Mg ++ , Zn ++ , and Ni+ 4 , 
which also produce passivity, are anodic to 
Sn and their influence cannot be explained 
by such reasoning. However, their passi- 
vating tendency is very weak compared 
with Cu++ and Ag + . In the other class of 
cations (normal) all the elements are 
anodic to Sn except Bi ++ *, and no film 
formation should be expected. 

Of all the elements, only zinc acting as an 
alloying agent and also as a cation in solu 
tion produces the same effect. With all 
others, no similar relationship exists. Such 
an element as Mg, which is far removed 
from Sn anodically, in the e.m.f. series 
induces passivity as Mg ++ but is harmful 
as an alloying element. Cu and Ag, 
which are cathodic to Sn, increase corrosion 
resistance as Ag + and Cu ++ but are not 
beneficial as alloying elements, because the 
ions do not enter the corroding medium; 
these metals are in the cathodic areas of the 
alloy, and cannot build up a protective 
film. 



208 



STUDIES UPON THE CORROSION OF TIN 



There are other factors that influence the 
corrosion properties of tin and its alloys, 
which have not been considered in this 
paper. Grain size 5 has been shown to have 
some effect on the amount of corrosion of 
tin in carbonate solutions; the finer the 
grains, the less the corrosion. However, 
since most of the alloying elements tend to 
reduce the grain size of tin, it is difficult to 
isolate the effect of grain size from the data 
presented. Other factors, such as solubility 
of the corrosion product and hydrogen 
over-voltage of the alloy phases, naturally 
should influence the corrosion characteris 
tics of tin in alkaline solutions. It will be 
necessary to obtain more experimental data 
before these factors can be discussed. 

SUMMARY 

j. Time-potential curves have been 
made for the following cations in sodium 
carbonate solutions of pH 11.2 and 8.4; K + , 



Li + , NH 4 + Ca++, Fe + +, Cd ++ , 

Bi + + + , Co++ Mg ++ , Cu++, Pb*+, Zn ++ , 

Ni+ + , Ag+, and In + ^ + . Of these, Mg++, 

Ag+, Pb+ + , Cu++ Zn++, Ni ++ promoted 

passivity. 

2. Arsenate, arsenite, silicate and sul 
phide ions have been added to sodium 
carbonate solutions of pH 11.2 and 10.0; 
silicate caused passivity in this range. 

3. The effect of oxygen and nitrogen on 
the corrosion of tin has been studied. 
Oxygen induces passivity, while nitrogen 
increases the corrosion rate. 

4. Measurements over the entire pH 
range of carbonate solutions show little 



difference between tin of very high purity 
and Chempur tin. 

5. Time-potential curves have been 
made for binary tin alloys containing the 
following elements; Cu, Zn, Ni, Fe, Pb, Ag, 
In, Mg, Cd, Sb, Bi, and Ca. Weight-loss 
measurements in selected cases, Cu, Zn, Sb, 
and In, alloyed with Sn were made, and 
measurements were of the order predicted 
by the time-potential curves. Zn and In 
were the most effective in reducing 
corrosion. 

6. Films developed by heating the vari 
ous alloys at temperatures of io5C. for 50 
min. and i75C. for 5 hr. provided no addi 
tional protection to the alloys in the 
alkaline solutions. 

ACKNOWLEDGMENT 

The Bristol-Myers Co., of New York 
City, and the Sun Tube Corporation, of 
Hillside, N. J., have sponsored the research 
program of which this investigation is a 
part. 

REFERENCES 

1. G. Derge: Studies upon the Corrosion of Tin, I 

Potential Measurements on High-purity Tin in 
Carbonate Solutions. Trans. A.I.M.E. (1938) 
128, 301, 

2. G. Derge and H. Markus: Studies upon the Cor 

rosion of Tin, II The Effect of Other Anions in 
Carbonate Solutions. Trans. A.I.M.E, (1939) 
*33 294. 

3. G. Derge and J. W. Stewart: Extrusion of Tin and 

Its Alloys. Trans. A.I.M.E. (1940) 137, 389. 

4. R. H. Brown, W. L. Fink f and M. S. e Hunter: 

Measurement of Irreversible Potentials as a 
Metallurgical Research Tool. This volume, 
page us* 

5. G. Derge: Relations between Crystal Structure and 

Corrosion. Trans. Elec. Soc. (1939) 75,440-62. 

6. D. Hanson and E. J. Sandford: Some Properties of 

Tin Containing Small Amounts of Aluminum, 
Manganese, or Bismuth. Jnl. Inst. Metals 
(i93S) 56, 43-58. 



Effect of Composition on Physical and Chemical Properties 
of 14-karat Gold Alloys 

BY TRACY C. JARRETT,* JUNIOR MEMBER A.I.M.E. 
(Cleveland Meeting, October 1940) 



IN i4-karat gold alloys, as in lo-karat 
gold alloys, 3 the addition of such metals as 
zinc, nickel, copper and silver produces a 
wide range of physical and chemical proper 
ties such as color, hardness, corrosion resist 
ance and melting points. These important 
factors made gold alloys desirable in the 
optical work and for jewelry. 

In i4-karat gold alloys, as well as in 
lo-karat or i2-karat, it is possible to have 
two golds of the same color, one of which 
will corrode more than the other, although 
it may be harder. Some of these alloys have 
been investigated by F. E. Carter, 1 E. M. 
Wise, 5 and L. Sterner Rainer. 4 

In this group of alloys the gold content of 
58.3 per cent (i4-karat) was held constant, 
while the copper varied from 36 to 24 per 
cent. The balance of the alloy consisted of 
one, two, or three of the metals silver, zinc, 
and nickel. 

All the alloys were melted in an induction 
furnace and cast into an ingot that meas 
ured 3 by K by % in. The method of 
determining the melting points, color deter 
minations and acid test have been discussed 
in previous papers by the author. 2 - 3 

It is necessary to use small test bars 
(Fig. i) . The tensile bars were annealed at 
i3ooF. for 20 min. and then air-cooled. 
There being a very marked drop when the 
yield point was reached, "the drop of the 
beam" method was used in determining 
the yield point. 



Manuscript received at the office of the Institute 
June 8, 1940. Issued as T.P. 1249 in METALS TECH 
NOLOGY, December 1940. 

* Assistant Metallurgist, American Optical Co., 
Southbridge, Mass, 

3 References are at the end of the paper. 



ALLOYS 

Gold- cop per -zinc Alloys. Gold-copper- 
zinc alloys melt, cast, and fabricate easily. 
The low melting points, however, limit 




FIG. i. TEST BAR. 

their use to some extent. The physical 
properties of these alloys are not entirely 
satisfactory. When the ultimate strength is 
satisfactory, the elongation and reduction 
of area are too low (Table i) . 

Gold-copper-zinc-nickel Alloys. In this 
group of alloys, in which zinc and nickel 
replace copper in approximately equal parts 
(Table i), the yield point, hardness, and 
ultimate strength increase with the increase 
of zinc and nickel content, while the reduc 
tion of area and elongation decrease. The 
physical properties are desirable, particu 
larly of the white gold alloys. The alloys of 
this group are not difficult to cast and 
fabricate, and most of these colors could be 
used commercially. 

Gold-copper-nickel A Hoys. Gold-copper- 
nickel alloys are similar to the gold-copper- 
zinc and the gold-copper-zinc-nickel alloys. 
They do, however, possess high yield point, 
ultimate strength and hardness, somewhat 
higher than those of the other two groups 
mentioned (Table i). The difficulty with 
these alloys is that they fire-crack readily 
and heavy rolling is necessary between 
annealings. Being hard to fabricate, many 



209 



210 EFFECT OF COMPOSITION ON PROPERTIES OF I4-KARAT ALLOYS 



1, 



Hardness. 


Rockwell 
B Scale 


*SSS 


mo o M m n 


1 


moo o> o w n 

irj 1/5 mO O O 


tt r- N moo o 
O O t r rco 


r- w r- o on 
r-co oo Ov o o 


c 
.2 

c 

2 


CJC 


m o o n -<t 


o o o c o o 


oo o in n 


o ^ n o o o 


OO OOO 


oooooo 


c 

c 


I 








^^^"^^ 






3 
". 


3 c 


o o in m o m 

O m r- M o r* 

OOO O -T n 


m m o o m c 

^ M o r^- o 

O O O O m m 


O in o m o O 

O r- o O o\ m 


o m m o O m 

o M M d : 

OOOOO m 


m m o mo m 
10 m m m m m 


o rf o o v> m 
r- m M m t-r- 


s" 


i|^" 


OOOOOO 

oooooo 
noo o o o c 


OOOOOO 

oooooo 


C O O O O 

oooooo 
o o n o o n 


OOOOOO 

oooooo 
o m o n o n 


oooooo 
oooooo 
o o noo N o 


o o o o o c 
oooooo 
oo oo n> o c 


~4 


3-^ 










O r- r-* r-oo co 


r-co co o\ H M 

H M 


^ 


i^t/D 


O C O O 

C C O 

q N + n m q 


C O O 

oooooo 
q q q M qo_ 


oooooo 
oooooo 

M in O M LTJ M 


oooooo 
oooooo 

O to ^ O OO 

in m <o oc"c< 


OOOOOO 

oooooo 
q M oo n q q 

CO M ro r rf O 


00000 
o o o o c 
OQ q q M H r- 

o" M oo" cTooo" 
Tj-mctr-oooo 


bo 

c 


-&; 


m o o m m o 


m o o o o m 


O O m o m o 


o m in m m m 


3 O O O O m 


n mmo o o u- 


*2 


5 ? 


O o O O m m j 


q r- t- r- i- r- r- 


00 CO 00 00 00 0\ 




jj t- r- r- t- r-o 




s 




s I 


2 HHMMMH i 

5 c 


J (. 
a 


4 

2 


^ 


5 


rrosion 

LnQtj 


t, i 

! l*i 


< 
A H 

v OOOOOO 2 

5 oooooo 


t 

J a 

; O C 
- OOOOOO % 


S t 

fl OOOOOO k 

4 oooooo H 


4 
j 
o n MOOCOO 

J O M M O O 

5 


j O O O O C 

i oooooo 


J 

d -<t <N OvOO CO CO 
J M M 

-> oooooo 

5 OOOOOO 


o 


o 3 

1 1. Jj 




- J 




j 


, 


^ oooooo 


tu 
O 


1 ! 

5 I 


i i 

s c 



t t- n mco i o\ " 
H M noo u, g 

p 


^ rtmmm o ^ 

3 o" oo" 6 N T}- r- c 


; 

i 


- o o n MOO o 
U N oo m n m 


5 O PC N r- N ,c- 

* r- M IN o ^O 

2 w -ri-m\o t-oo 


u, 

o 


o ^- 


/ O * oooo r}" C 
c?ioQ OQO noo _ 


<t ooo n M o\ c 
a o w M moo ^ 


i 

5 y 


too oo r-oo 


j in N t> i Ov O 
* co Ovoo o> moo 




* 





mcO O N rj-o f^ 


-tmo r-r- 


3 




] ~w*> 




*O 

C 


t-, 

O, 

a 


i-oc 5i o noo 

m CH o a r- -=t 


m m SN H n cs 

in n w o\ r m 


J* MM SS 


oo m r- m -*f o 

(N Tt- N H O\ M 


in r- c O N o 

o <N o m o a 

m IN M co r- ^ 


& n t o *f T 
O r-m t-i o\oo 

in n M oo T 
n n n w M 




1 








in IN O\o0 O <N 


M OO 1-1 

o t- n o\ r- m 


m rr H ro if o 
NOO Ov o\ n H 




















*o 
O 


00 OO 00 CO 00 00 

in in n m m m 


00 COCO CO CO CO 


oo co oo oo coco 
mmm m m m 


00 OO CO 00 CO 00 

m m in in m m 


m in m in m m 


CO CO CO COCO CO 

in m m m m m 


V 

J 

* 


3 








::::.: 










^^^O^^O^ 

U J-H t-, !_ _ 


Sllll 


S sis 


Sllill 


liil! 


. . o 


j: 


I 


no 0,2^00 


1-1 (-( M 


-^0* 


w ^ r o no 


o no Ov m 
N c* M N n n 


O\ <N moo M Tt 
M N N M con 



TRACY C, JARRETT 



211 



of them cannot be used as commercial 
golds. 

Gold-silver-copper-zinc Alloys. In this 
group of alloys, in which the silver and zinc 
are present in approximately equal propor 
tions, the physical properties show little 
change as the percentage of silver and zinc 
increases (Table i). The physical properties 
in this group are more desirable than those 
of the gold-copper-zinc alloys. These alloys 
process easily and may be used as com 
mercial gold. 

Gold-silver-copper -zinc-nickel A lloys. In 
this group of alloys, with three elements 
replacing copper in approximately equal 
proportions, the yield point and ultimate 
strength increase as the content of silver, 
zinc and nickel increases. Reduction of area 
remains fairly constant with the increase of 
silver, zinc and nickel, while the elongation 
decreases after 12 per cent addition of 
silver, zinc and nickel has been reached 
(Table i). These alloys, possessing good 
physical properties, fabricate easily, mak 
ing them usable as commercial gold alloys. 

Gold-silver-copper-nickel Alloys. The 
yield point, hardness and ultimate strength 
increase in gold-silver-copper-nickel alloys 
as the silver and nickel content increases, 
while the elongation and reduction of area 
decrease (Table i). A few of these alloys 
may be used commercially, but most of 
them are difficult to cast and fabricate. 

DISCUSSION OF PROPERTIES 

The presence of silver in the i4-karat 
alloys brings about a slight change in corro 
sion loss (Table i). The total corrosion loss 
for similar alloys in the i4-karat golds is far 



less than for the io-karat gold alloys. 3 This 
change is brought about by the increase of 
gold content and was expected, as it is well 
established that the corrosion loss decreases 
as the gold content increases. The effect of 
silver in the alloy, however, is not marked 
in the i4-karat gold alloys as in the ic-karat 
groups, and this may be owing to the pres 
ence of a greater gold content. 

Nickel plays the same part in the 14- 
karat gold alloys as it does in the io-karat; 3 
that is, it helps to reduce corrosion loss 
and increases the hardness, yield point and 
tensile strength. Zinc acts in just the oppo 
site way in these alloys, as it reduces 
the hardness, yield point and ultimate 
strength. 

The effects of silver, zinc, and nickel are 
not as noticeable in the i4-karat gold alloys 
as in the io-karat series. Corrosion loss is 
very low and the resistance to hydrogen 
sulphide is far greater than in 10 -karat 
golds. 

This group of alloys does not have as 
many marked changes as the io-karat 
group. It is known that the intermetallic 
compounds formed in these alloys have an 
important influence on the properties of the 
alloys. 

REFERENCES 

1. P. E. Carter: Gold, Silver, Copper Alloys. Trans. 

A.I.M.E. (1928) 78, 786. 

2. T. C. Jarrett: Effect of Composition on Color and 

Melting Point of io-karat, 12-karat and 14-karat 
Gold Alloys. Trans. A.I.M.E. (1940) 137, 45.6. 

3. T. C. Jarrett: Effect of Composition upon Physical 

and Chemical Properties of io-karat Gold 
Alloys. Trans. A.I.M.E. (1940) 137, 447- 

4. L. Sterner- Rainer: Einige Eigenschaften der 

Legierungen Au-Ag-Ctt. Ztsch. Metallkunde 
(1926) 18, 143. 

5. E. M. Wise: High-strength Gold Alloys for 

Jewelry, and Age-hardening Phenomena in Gold 
Alloys. Trans. A.I.M.E. (1929) 83, 384. 



Beneficial Effects of Zirconium in Cast Nickel-silicon Bronzes 



BY F. R. HENSEL,* MEMBER A.I.M.E., E, I. LARSEN,* AND A. S. DOTY* 

(Chicago Meeting, October 1939) 

THE alloy under discussion is a Fig. i shows results of heat-treating tests 

nickel-silicon bronze and is one of many on two typical nickel-silicon bronze castings 

age-hardening or precipitation-hardening of the type on which the observations were 

TO, ^ ., 




Quenched 
from 750 "C. 
Aged af4$0C. 




Ouenched 
from850C. 
Aged at 450 C. 





1 2 4 8 163264. 1248 163264 T 248 163264 
Aging t i me .hours 

:, /. WNi, OL44Si, 0.046Fe -Ni/Si = 4.52/1 
:, 0.043Fe -Ni/Si =3.66/1 



Quenched 
from750C, 
Aged at 500*0. 



Quenched 
from 65Q C C, 
Aged 0+500*0. 




Quenched 

from90Q*C. 

Agedc*+5QO*C. 



_J L_ 



_! L_ 



FIG. i.- 



2 4 8 163264 1248 163264 I 2 4 8 163264 
Aging time , hours 

-EFFECT OF QUENCHING AND AGING TEMPERATURES ON HARDNESS OF TWO TYPICAL NICKEL- 
SILICON BRONZES, NO. 2351 AND No. 2362. 



copper-base alloys. The hardening con 
stituent is a nickel-silicon compound. 1 The 
quenching temperature necessary to pro 
duce hardening by a subsequent aging 
treatment is a function of the composition 
and extends from about 7ooC. to Q5oC.; 
the aging temperature range extends from 
400 to 6ooC. 

The work carried out by the authors of 
this paper was confined to cast alloys 
having compositions within the ranges of: 
1.5 to 3.00 per cent Ni; 0.4 to 0.75 per cent 
Si; balance, copper, 

Manuscript received at the office of the Institute 
Oct. 21, 1039. Issued as T.P. 1237 in METALS TECH 
NOLOGY, December 1940. 

* P. R. Mallory and Co., Indianapolis, Ind. 

i References are at the end of the paper. 



made. A quenching temperature of 75oC. 
produces almost as high a hardness on 
subsequent aging as do quenching tem 
peratures of 850 and 900 C. 

Several foundries have spent years of 
active and intelligent effort in trying 
to control the casting and heat-treat 
ing variables of nickel-silicon bronzes and 
finally gave up the attempts because of 
the inconsistencies of results. 2 This unde 
sirable characteristic was not manifest in 
the ultimate hardness or conductivity 
but primarily in tensile strength and 
ductility, as shown in Table i. The vari 
ation in tensile properties was often much 
greater than that existing on the alloys 



F. R. HENSEL, E. I. LARSEN AND A. S. DOTY 



213 



of Table i This table is shown because 
these eight test bars were from successive 
heats of the alloy made tinder conditions 
duplicated as closely as possible by the 
best modern foundry practice. In some 



comparison with tests on straight nickel- 
silicon bronze castings, which showed an 
average of only about 50,000 Ib. per sq. in., 
with probably 35 per cent of the test bars 



instances test bars actually broke during u 

machining. 

70,000 

TABLE i. Characteristics of Cast Nickel- 
silicon Bronzes 5.60000 






Ni- 




to/7, 


em 


thO. 


?5% 

*-* - 


Zrc 

= 


ride 

sr"-< 


i 

,--*^ 


- 










X 




^ 








^* 
















^ 


/ 




Specimen 
No. 


Hardness, 

Rockwell 
B 


Electric Con 
ductivity, 
Per Cent 
I.A.C.S. 


String. |SO,000 
per Sq. In. ^ 

c 














f 


x 










2646 
2647 
2648 
2649 
2660 
2661 
2662 
2663 


89 

90 
94 
96 
89 
90 


39-2 

40. 1 
39-7 
39 7 
38.0 
37-7 
38.6 
38.3 


55 40,uOO 
23,800 jy 
29,200 *<2 
23,800 ^^nnnn 

51,200 7;W,UUU 
45,800 "o 
48,200 E 

li^o i 20 - 000 












; 























A 


w g 


hth 


i-S 


Bn 


>/7Z 






























10000 
In an effort to trace the cause of the 

trouble, various melting and casting tech- 



























niques were employed. The Ni2Si content 
was varied from 1.5 to 5 per cent and the 
nickel-silicon ratio varied from 3 to i to 
6.5 to i. No appreciable advantage or 
disadvantage was noted. In order to elimi 
nate the trouble, additions were made of 
magnesium, beryllium, cadmium, zinc, 
manganese, phosphorus, iron, lithium, ti 
tanium, thorium, uranium and zirconium. 
Most of the additions were below 0.25 per 
cent. Certain single tests, such as 0.25 per 
cent Cd, 0.25 per cent Mg, 0.25 per cent Th 
and o.i per cent Mn, indicated some im 
provement, but continued investigation of 
the alloys still showed inconsistencies in 
regard to tensile strength. 

Out of all the elements added in an 
attempt to improve the tensile strength 
of nickel-silicon bronze castings, zirconium 
alone gave consistently effective results. 
Test bars from at least 20 successive pro 
duction heats to which approximately 0.25 
per cent Zr was added showed no ulti 
mate tensile strength below 56,000 Ib. per 
sq. in. The average was about 65,000 
and some tests showed over 70,000. In 



I "2 3 4 5 6 7 8 9 10 11 12 

Test number 

FIG. 2. TENSILE STRENGTH OF TEST BARS 
FROM FIRST TWELVE PRODUCTION KEATS WITH 
0.25 PER CENT ZIRCONIUM ADDED, COMPARED 
WITH TWELVE SUCCESSIVE HEATS OF NICKEL- 
SILICON BRONZE. 

having less than 35,000 Ib. per sq. in. 
ultimate tensile strength, a remarkable ad 
vantage was noted in the addition of 
zirconium. 

Fig. 2 gives a graphic illustration of the 
tensile strength of test bars from the first 
12 production heats to which 0.25 per cent 
Zr was added as compared with 12 suc 
cessive heats of straight nickel-silicon 
bronze. The test results are arranged in 
order of increasing tensile strengths. This 
graph clearly shows the distinct improve 
ment brought about by the additions of 
zirconium. Not only is the average strength 
considerably higher, but, more important, 
there are no test bars showing a tensile 
strength below 55,000 Ib. per sq. in., 
whereas about 35 per cent of the tests on 
straight nickel-silicon bronzes show an 
ultimate tensile strength below 35,000 Ib. 



2I 4 



ZIRCONIUM IN CAST NICKEL-SILICON BRONZES 



per sq. in. and only 15 per cent above 
55,000 Ib. per sq. inch. 



strengths. Variation of the zirconium con 
tent from as little as o.i per cent up to 




FIG. 3. FIG. 4. 

FIG. 3. STRAIGHT NICKEL-SILICON BRONZE AFTER QUENCHING FROM 9ooC. AND AGING 16 HOCRS 

AT 45C. X 75- 

Note segregation in grain boundaries. Sample was polished, then etched in NH^OH -f- H^Oz. 
FIG. 4. NICKEL-SILICON BRONZE CONTAINING 0.28 PER CENT ZIRCONIUM AFTER SAME TREATMENT. 

X 75- 

Although some segregation is evident, it is not continuous nor nearly as pronounced as in 
Fig. 3. Sample was polished, then etched in NH^OH -f- H 2 O 2 . 




A / 2 I 2 4 5 16 32 
kquenched A 9 n ^ time,hours 

FIG. 5. ELECTRICAL CONDUCTIVITY AND HARD 
NESS VERSUS AGING TIME. 
Sample L-894-A quenched from 9ooC., aged 
at 45oC, Nickel-silicon bronze with 0.25 per 
cent Zr added. Sand-cast. 

Electrical conductivity sample, ?|-in. diam 
eter by 6 in.; hardness sample, i by i by }^ 
inch. 

In addition to test bars from production 
heats, several test bars from small experi 
mental melts showed equally good tensile 



70,000 
60,000 

. 50,000 
o* 

.0 
i 30,000 

^20,000 
10,000 














0.2% 
61,000 


f e/ctpo 
Ikperst 


J.ff7. "\ 










/ 


^T 










^-Prop 
4S.C 


vrfiona! limit 
OOlbpersq./n. 








/ 












/ 












/ 














/ 














0.002 0004 _ 0.006 _ 0.008 0010 0.02 0,0 



v wwf U.UUO UUUO U UIU 

Strain, inches in 2-inch length 
FIG. 6. STRESS VERSUS STRAIN. 
Nickel-silicon bronze with 0.25 per cent Zr 
added. Sand-cast. Ultimate tensile strength, 
68,550 Ib. per sq. in. Sample L-QI 7-1. 

i per cent did not affect the results. Figs. 3 
and 4 offer a possible explanation of the 
beneficial effects of zirconium. 



F. R. HENSEL, E. I. LARSEN AND A. S. DOTY 



215 



In all tests with additions from 0.08 per 
cent up to 0,73 per cent Zr the grain 
boundaries were relatively fine and free 
from excessive segregation. 

The addition of zirconium had no ill 
effects on the hardness and electrical con 
ductivity of the alloys. 

Fig. 5 shows typical aging curves on a 
nickel-silicon bronze to which 0.25 per 
cent Zr was added. Hardness reaches a 
maximum of approximately 90 Rockwell B 
after 16 hr. at 45oC. and conductivity 
reaches approximately 37 per cent I.A.C.S. 
after the same length of time. Fig. 6 shows 
a stress-strain curve on the same alloy. 
Ultimate tensile strength was 68,550 Ib. per 
sq. in.; yield strength at 0.2 per cent 
elongation, 61,000 Ib. per sq. in.; propor 
tional limit, 45,000 Ib. per sq. inch. 

The information presented in this paper 
is primarily a resume of factual data. The 
research work concerning the theoretical 
explanations is not complete. As mentioned 
before, the microstructure of the straight 
nickel-silicon bronzes indicates that the 
extreme brittleness is in some way con 
nected with the type and location of the 
precipitate found during aging after 
quenching. Possibly the presence of zir 
conium which in itself, or in combination 
with another element, can be used as a 
precipitation-hardening constituent in cop 



per may alter the precipitate (either in 
respect to type or location) sufficiently to 
eliminate the intercrystalline weakness and 
general embrittlement found in cast nickel- 
silicon bronzes. Furthermore, zirconium is 
a most efficient degasifying agent, pro 
ducing castings free from gases usually 
occluded within the grain boundaries. It 
seems quite possible that occluded gases 
may be the cause of incipient cracks during 
heat-treating. Another advantage of the 
zirconium addition was the effect on the 
formation of nuclei by zirconia, resulting in 
a finer grain structure, 

REFERENCES 

1. M. G. Corson: Copper-alloy Systems with Vari 

able Alpha Range, and Their Use in the Harden 
ing of Copper. Proc. Inst. Metals Div., A.I.M.E. 

2. H, W. Gillett: The Role of Silicon in Non-ferrous 

Castings. Amer. Foundrymen s Assn. Preprint 
No. 38-10. 

3. H. W. Bassett: Copper and Copper Alloys. Uin. 

and Met. (April 1928) 9, 170. 

4. J. L. Gregg: Dispersion-hardening in Copper-base 

and Silver-base Alloys. Trans. A.I.M.E. (1929) 
J$3 409. 

5. W. C. Ellis and E. E. Schumacher: Heat-treat 

ment and Mechanical Properties of Some Cop 
per-zinc and Copper-tin Alloys Containing 
Nickel and Silicon. Trans. A.I.M.E. (1929) 83, 

6. W. C." Ellis and E. E. Schumacher: Effect of Com 

binations of Strain and Heat-treatment on 
Properties of Some Age-hardening Copper 
Alloys. Trans. A.I.M.E. (1931) 93* 373- 

7. C, L. Wilson, H. F. SilHman and E. C. Little: Rate 

of Precipitation of Nickel t Silicide in the 
Hardenable Copper- nickel- silicon and Copper- 
cobalt-silicon Alloys. A.I.M.E. Contribution 11 

8. B. W. Gonser and L. R. van "Wert: The Age-hard 

ening Characteristics of Some Copper-nickel- 
silicon Alloys. Metals and Alloys (Nov. and 
Dec. 1934). 



Study of the Metallography and Certain Physical Properties 
of Some Alloys of Cobalt, Iron, and Titanium 

BY CHARLES R. AUSTIN* AND CARL H. SAMANS,* MEMBER A.I.M.E. 



(Cleveland Meeting, October 1940) 



IT has been known for several years 1 that 
certain alloys of the Konal type, containing 
commercial cobalt (99.32 per cent Co and 
0.42 per cent Ni) and varying amounts of 




PERCENT Fe^Ti 

FlG.tl. 

FIG. i. EFFECT OF VARYING AMOUNTS OF 

FERROTITANIUM (FEjTl) ON PHYSICAL PROPER 
TIES OF COMMERCIAL COBALT. (C. R. Austin. 1 ) 

Upper section. Tested at 6ooC. after water- 
quenching from 95oC. 

Lower section. Tested at 6ooC. after water- 
quenching from 95oC. and aging 72 hr. at 
6 5 oC. 

FIG. 2. PORTION OF QUASI-BINARY DIAGRAM 
COBALT-FEgTl AS DETERMINED METALLO GRAPH 
ICALLY. 

ferrotitanium, exhibit very high tensile 
strengths at 6ooC. However, their pro 
portional limits and yield points are rela 
tively low. Published data 1 for the alloys 
quenched in water from 95oC. and then 



Manuscript received at the office of the Institute 
Dec. i, 1939; revised Oct. s, 1940. Issued asT.P. 1257 
in METALS TECHNOLOGY, December 1940. 

* Professor and Associate Professor of Metallurgy, 
respectively, The Pennsylvania State College, State 
College, Pa. 

1 References are at the end of the paper. 



tested at 600 C. are summarized in the 
upper section of Fig. i. Similar specimens 
that were aged at 65oC. for 72 hr., after 
quench and prior to testing at 6ooC., gave 
the results shown in the lower section of 
Fig. i. 

TABLE i. Analysis of Material 



Constituent 


Percentages 


Sample 
3125 


Sample 
3124 


Sample 
3123 


Si 


0.25 
4-15 
1. 15 
0.30 
94-15 


0.51 
10. 70 
3-58 
0.41 
84.80 


0.60 
I3-I7 
4-70 
0.42 

81 . ii 


Fe 


Ti 


Mn 


Co (by difference) 





a The cobalt from all three samples, united and 
tested for nickel, gave 0.54 per cent Ni based on the 
weight of cobalt. 

In the present work the investigation of 
the properties of these materials has been 
extended by a study of the tensile deforma 
tion characteristics in moderately long 
time tests at 600, 700 and 8ooC. after 
quenching from various solution tempera 
tures. In addition, some metallographic 
studies have been made to correlate with 
the deformation tests, and to obtain in 
formation on the quasi-binary system 
cobalt-Fe 2 Ti. 

The material, very kindly furnished by 
the Westinghouse Electric and Manufac 
turing Co. in the form of hot-swaged rods 
of %6-i n - diameter, analyzed as shown in 
Table i. 

The alloys, when originally made up, 1 
were based upon an iron-titanium com 
pound, FesTi, as indicated by the work of 



216 



CHARLES R. AUSTIN AND CARL H. SAMANS 



217 



J. Lamort, 2 W. Koester and W. GeUer, 3 
and W. Jellinghaus, 4 The more recent 
studies of H. Witte and H. J. Wallbaum 6 
have shown that the compound is probably 
Fe 2 TL* 

QUASI-BINARY SYSTEM CoBALT-FE 2 Ti 

In order to determine the nature of the 
system Cobalt-Fe2Ti more completely, 
the three alloys available were examined 
metallographically after quenching from 
various temperatures. Although the deter 
mination of true precipitate by this method 
was difficult because of the numerous inclu 
sions, probably titanium oxide and nitride, 
reasonable data were secured by directing 
attention to the grain boundaries and 
assuming that as long as they were clean, 
no particles of a second phase existed. On 
this basis concordant data as shown in the 
diagram of Fig. 2 were derived. Extrapola 
tion of the solidus and solvusf lines indi 
cates a eutectic temperature at about 
i2ooC. and a maximum solubility of about 
25 per cent Fe 2 Ti in cobalt. 

PRECIPITATION-HARDENING 
CHARACTERISTICS 

From the slope of the solvus line the 
alloys would be expected to be susceptible 
to precipitation-hardening. In investigating 
this phase of the subject the three alloys 
were quenched after 15 min. at one of three 
temperatures: 950, 1150, and i3ooC., 
and then aged for times up to 100 hr. at 
600, 700, 8oo ? and gooC. The diagram 
in Fig. 2 shows that for the alloys 3124 and 
3123 the lowest "solution" treatment did 
not even approximate complete solution, 





3125 


3124 


3123 


Fe 


1.51 


2.38 


2 . 25 


Fe*Ti 


3 79 


II .90 


15.62 











t Solvus line is the line defining the limit of 
the primary alpha solid solution with respect 
to temperature and composition; sometimes 
called the solid solubility curve. 



while the highest treatment was just below 
the solidus. Also, it is evident that for alloy 
3125 no hardening would be expected for 
the two higher " precipitation " treatments, 
as the alloy would still be in the solid 
solution condition. 

An examination of the data on the three 
alloys, shown in Fig. 3 with Rockwell B 
hardness as ordinate plotted against aging 
time as abscissa, discloses two discordant 
facts, both in connection with the low 
ferro titanium alloy, No. 3125. 

1. The higher the solution temperature, 
the lower is the as-quenched hardness, and 
the magnitude of this difference (about 28 
Rockwell points) is greater than would be 
anticipated from grain coarsening alone. 

2. Appreciable hardening resulted from 
aging the quenched alloy at 800 and 
QooC., even though, for these tempera 
tures, the alloy falls completely within the 
solid solution field (Fig. 2) . The magnitude 
of the hardness increase resulting from this 
varied, being about 20 points for the 
i3ooC. treatment, about 10 points for 
that at nooC. and only 3 points for that 
at o5oC. 

The first of these effects could be caused 
either by plastic deformation during 
quenching or by a partial transformation 
from the face-centered cubic beta phase, 
stable at the high temperatures, to the 
close-packed hexagonal alpha (or epsilon) 
phase, stable below about 45oC. All the 
quenched specimens of this alloy show 
numerous stria tions (Fig. 10) under the 
microscope and these could readily be ex 
plained by either of the two reasons 
suggested. 

The age-hardening in an apparently uni 
form solid solution alloy, on the other hand, 
probably results from the effects of a strain, 
which might be produced by either of the 
reasons cited, and thus is similar to the 
phenomenon of double aging, which has 
been so extensively studied by R, H. 
Harrington. 6 * As he points out, "an alloy 
normally constituted of a single solid solu- 



2l8 



STUDY OF SOME ALLOYS OF COBALT, IRON, AND TITANIUM 



tion phase in the annealed state may be 
truly precipitation-hardened" by a process 
of strain-aging in which recrystallization 
nuclei produce effects similar to those of 

3l25 



LONG-TIME TENSILE DEFORMATION 

CHARACTERISTICS 

For studying the tensile deformation 
characteristics under dead loading over 



3124 



3123 



100 




20 40 60 
AGING TIME 1N 

FIG. 3. CHANGE IN ROCKWELL B HARDNESS WITH TIME or AGING AT 600, 700, 800 AND gooC. 
Alloys 3125, 3124, 3123 quenched originally from 1300, 1150, and 95oC. 

precipitation nuclei in the usual age- 
hardening processes. 

If this strain-aging hypothesis is ac 
cepted, it follows that the higher the 
quenching temperature, the greater the 
strain induced and the coarser the grain 
size, and, consequently, the greater the rela 
tive effects of strain-aging when reheated. 
Apparently, the hardening of the fine 
grained material quenched from Q5oC. is 
completed during the quench while that of 
the coarse-grained structures requires sub 
sequent aging for its optimum develop 
ment. Because of the relatively low 
solubility of Fe 2 Ti in cobalt at 600 and 
7ooC. it is entirely possible that a large 
portion of the hardening found on aging at 
these temperatures is attributable to the 
same cause, although there are so many 
effects possible, such as order-disorder 
changes for example, that definite explana 
tions cannot be given. 



periods of several hundred hours, the spring 
loading furnace, which has already been 
described in the literature, 7 was used. This 
equipment tests a specimen having a gauge 
length 4 in. long and J in. in diameter 
(Fig. 9). Within the limits of the test, the 
smallest secondary creep rate that can be 
determined definitely is o.oi per cent per 
loco hr., and it will be noted that several 
of the rates were even less than this. Step 
loading was used throughout in increments 
of 2000 Ib. per sq. in., a given loading being 
maintained until the secondary creep rate 
became essentially constant. This condition 
generally required upward of 300 hr. The 
secondary creep rate, so secured, was then 
designated as the "deformation rate" char 
acteristic of the alloy and the applied stress. 
Previous work 7 has indicated that virtually 
the same secondary creep rates obtain 
regardless of whether the specimens are 
step-loaded or loaded throughout the test 



CHARLES R. AUSTIN AND CARL H. SAMANS 



219 



with a constant load, always provided, 
naturally, that the rates remain reasonably 
low so that the section does not decrease 
markedly. As a comparative method the 
test has already proved its worth. However, 
these creep data are neither presented nor 
intended for use for purposes of engineering 
design, but rather as reasonably accurate 
results of the reactions of these alloys to the 
specific conditions studied. An additional 



before testing in order to obtain an approxi 
mately similar structural state. Tests were 
run at 600, 700, and 8ooC. and complete 
data on these tests are compiled in Table 2. 
In Fig. 4 the deformation curves for the 
tests at 600 C. are shown.* Alloy 3125 
is the only one that does not show negative 
creep after the application of the initial 
load, and this behavior is accounted for by 
the fact that it is only slightly super- 



TABLE 2. Tensile Deformation Characteristics at 600, 700 and 8ooC. after Air- 
quenching from 15 Minutes in Hydrogen at 95oC. 
PERCENTAGE PER 1000 HOURS 









Load, Lb. per Sq. In. 


eu 






2000 


4000 


6000 


8000 


10,000 


1 




< 

(U 


1 


fil 


C 

.2 


1 


g| 


c 

o 


oJ 
O 


1g 

?l 





I 


H 





C!J 

O 


ll 


C 

.2 


1* 


o 


PH 
H" 


oS 

1 


e 


0) 


03 



11 




urs at 




O +" 


e 


ffi ^ 
ft 

si 


. 

"^ 


U3 


al 


g q 














O 






o 




CJiv 1 


o 






O 






H 


< 





W 


o 


Q^ 


w 


o 


T 




u 


Q^ 


w 


L 


Q 


W 


O 4 


Q 


600 


3125 


3-79 


500 


500 


<O.OI 


502 


1002 


<O.OI 


502 


1504 


0.03 


505 


2009 


0.05 


616 


2625 


0.03 




3124 


11.90 


500 


500 


O.OI 


502 


1002 


O.OI 


502 


1504 


0.03 


505 


2009 


0.05 


616 


2625 


o. 05 




3123 


15.62 


342 


342 


<O.OI 


328 


670 


0.16 


297 


Q6 7 


0.22 


570 


1537 


0.31 










2867 


? 


500 


500 


<O.OI 


502 


1002 


0.03 


502 


1504 


0.04 


505 


2009 


0.08 


616 


2625 


o. 04 


700 


3125 

3124 


3-79 
11.90 


562 
562 


562 
562 


0.05 
0.06 


1178 
1178 


1740 
1740 


0-35 
0.04 


861 
86 1 


2601 
2601 


0.08 


503F 
579 


3104 
3180 


o. 15 


569 


3749 


0.70 




3123 


15.62 


263 


263 


0-53 


209 


472 


0.58 


209 


681 


0.6l 


238 


910 


88 










2867 


? 


562 


562 


O.II 


1178 


1740 


0.38 


86 1 


2601 


o 58 


579 


3180 


1.20 


569 


3749 


2.15 


800 


3125 


3-79 


524 


524 


5-75 


244 


768 


16.9 






















3124 


11.90 


524 


524 


0.63 


244 


768 


3-6 






















3123 


15.62 


210 


210 


0-43 


210 


420 


1-7 


209 


629 


8.6 















a This value is only an approximation, as the specimen fractured after 503 hr. at 8000 lb. per sq. in. without 
attaining a constant rate. 

saturated with ferrotitanium, as indicated 
in Fig. 2. In general, at this temperature 
alloy 3123, with the highest ferrotitanium 
content, seems to be inferior while there is 
little to choose between the other three. 

At the higher temperature of 7ooC., the 
secondary deformation rates were appre 
ciably greater (Fig. 5), and only alloy 2867 
showed any negative creep. This was of low 
magnitude and persisted only during the 

* It should be noted that, even though each 
loading is plotted as starting at zero time, the 
times at test temperature actually are cumu 
lative, and that this method of plotting was 
used only to obtain a clearer comparison be 
tween the various loadings. 



alloy, No. 2867, was included in these 
studies, although, because of insufficient 
material, it was tested only at 600 and 
700 C. As can be seen from its analysis 
(Ni, 27.0 per cent; Si, 0.71; Fe, 13.55; Ti, 
3.45; Mn, 0.28; Co, 54.05; with Al, 0.93 and 
C, 0.12 also reported), it is similar to alloy 
3124 with about 30 per cent of the cobalt 
in that alloy replaced by nickel and iron. 
Alloys 3125, 3124 and 2867 were tested 
at the same time while 3123 was tested 
earlier and reported erroneously in the 
literature 7 as alloy 3124. 

JA11 alloys were heated for 15 min. in 
hydrogen at 95oC. and air-quenched 



22O 



STUDY OF SOME ALLOYS OF COBALT, IRON, AND TITANIUM 



first few hours of test. Deterioration in note, too, that, except at the higher loads, 

3125 with increase of load may also be the substitution of nickel for cobalt (i.e., 

noted. This would seem to indicate a rela- 2867 compared with 3124), apparenlty has 

tively rapid removal of any hardening not been beneficial. 



;ooe 



j005 



r O04- 



003 



z 

tu 

002- 



5001 



TESTED AT 600!C 



4000 RSJ. 



20OO RSI. 






r 



200 4-00 



200 400 

TtKS 



CREEP RATE IN 
% PER IOOOHOUFT1 




y200 400 200 400 
RS ... 



FIG. 4. DEFORMATION CHARACTERISTICS OF ALLOYS 3125, 3124, 3123 AND 2867 AT 6ooC. UNDER 

SUCCESSIVE LOADS OF 2OOO, 4000, 6000, SOOO AND IO,OOO POUNDS PER SQUARE INCH. 




FIG. 5. DEFORMATION CHARACTERISTICS or ALLOYS 3125, 3124, 3123 AND 2867 AT 7ooC. UNDER 

SUCCESSIVE LOADS OF 2OOO, 4000, 6OOO, SoOO AND IO,OOO POUNDS PER SQUARE INCH. 



effects resulting either from the small 
amount of precipitate present or from the 
strains induced by quenching from 95oC., 
or by the deformation during the test. The 
relative weakness of 3123 still persists, and 
alloy 3124 is clearly the most resistant to 
deformation at this temperature in spite 
of the fact that 3123 should have appre 
ciably more precipitate available for 
strengthening purposes. It is interesting to 



The tests at 8ooC. J the results of which 
are plotted in Fig. 6, were not extensive. 
The three Cobalt-Fe 2 Ti alloys displayed 
high deformation rates, and 2867 was not 
tested because of insufficient material. 
Resistance to deformation at this tempera 
ture appears to be a direct function of iron- 
titanium additions. 

In Fig. 7 the results of the deformation 
studies are summarized in three plots of the 



CHARLES R. AUSTIN AND CARL H. SAMANS 



221 



logarithm of tensile deformation rate (i.e., 
secondary creep rate), in percentage per 
1000 hr., vs. the percentage of Fe 2 Ti. This 
indicates that the higher the test tempera 
ture, the greater the amount of Fe 2 Ti 
needed for improved resistance to deforma 
tion. There is no clear reason why the high- 
ferrotitanium alloy should be so much 
weaker at 6ooC. and at 700 C. than the 
one containing the intermediate amount. 
At the high temperature it would be ex 
pected to be stronger, and the results bear 
this out. Alloy 3125 also behaves according 
to expectation since, as Fig. 2 indicates, it 
would be essentially a solid solution at the 
two higher temperatures, and consequently 
would be expected to be weaker. A certain 
amount of information regarding the defor 
mation can also be secured from Fig. 8, in 
which stress is plotted against the logarithm 
of the deformation rate in percentage per 
1000 hr. For this method of plotting, the 
curves for all three alloys at 600 C., and 
for 3124 and 3123 at 7ooC., seem to be 
very steep near the inflection point. Other 



work on deformation in this laboratory has 
indicated that this shape curve is indicative 
of work-hardening during the deformation 




FIG. 6. DEFORMATION CHARACTERISTICS OF 

ALLOYS 3125, 3124, AND 3123 AT 8ooC- UNDER 
SUCCESSIVE LOADS OF 2OOO, 4OOO, AND 6OOO 
POUNDS PER SQUARE INCH. 

test. This would explain then, the weaken 
ing of 3125 at 700 and^8ooC., since its 




4 8 12 16 "0001 0.01 0.1 10 10.0 

PERCENT FE>Ti LOG TENSILE DEFORMATION RATE IN PERCENT PER 1000 HR3. 

FIG. 7. FIG. 8. 

FIG. 7. EFFECTS OF VARYING AMOUNTS OF FERROTITANIUM ON SECONDARY DEFORMATION RATES AT 

600, 700 AND 800C. UNDER LOADS OF 2OOO TO IO,OOO POUNDS PER SQUARE INCH. 
FIG. 8. VARIATION OF SECONDARY DEFORMATION RATES WITH STRESS FOR ALLOYS 3125, 3124 AND 

3123 AT 600, 700 AND 8ooC. 
(Curves with a steep inflection region indicate deformation in strain-hardening range.) 



222 



STUDY OF SOME ALLOYS OF COBALT, IRON, AND TITANIUM 



"recrystallization" temperature appears to 
lie between 600 and 7ooC.* However, 
inasmuch as the factors determining the 
creep rate at which the marked increase 

TOP 



HOT 
SHANK 



GAUGE 



FIG. 9. TYPICAL DEFORMATION SPECIMEN 
SHOWING LOCATIONS STUDIED METALLOGRAPH- 

ICALLY. 

in stress occurs are not as yet known, no 
definite information can be secured here to 
explain the apparently anomalous behavior 

of 3123. 

METALLOGRAPHY OF DEFORMATION 

SPECIMENS 

All of the deformation specimens were 
carefully examined metallographically in 



an effort to find correlations between the 
deformation characteristics and the micro- 
structure. The solution of 60 c.c. hydro 
chloric, 15 c.c. nitric, and 15 c.c. acetic acids, 
and 15 c.c. water recommended by Austin 
and Nickol 7 was found to be satisfactory 
for etching, although it was necessary to 
"age" the reagent before use. This solution 
was also used for electrolytic etching. 

Specimens from both the hot shank and 
the gauge section were cut as indicated in 
Fig. 9, so that transverse and longitudinal 
sections could be examined. In the tests 
at 8ooC. a specimen of some of the mate 
rials was held unstressed in the furnace for 
further comparison. 

Figs. 10 to 13 show the appearance of the 
four alloys after air-quenching from a 15- 
min. normalizing treatment at Q5oC. in 
hydrogen. All sections show evidence of 
nonmetallic inclusions, probably titanium 
oxide and nitride, but the structures appear 
to be essentially solid solutions. Evidence 
of a precipitate phase, however, may be 
noted in the photomicrographs of 3124 
and 3123, containing 11.90 and 15.62 per 
cent Fe2Ti, respectively (Figs, n and 12), 
In the alloy containing some nickel, a 
spheroidized phase, which may be the 
nickel-titanium compound NisTi (Fig. 13) 
is distributed. The extremely fine particles 
appearing at the grain boundaries appear 
to be similar in type to the precipitate 
found in alloy 3124 (Fig. n), although this 
was not established definitely. 

The normalized structures as modified by 
several hundred hours under stress at 
6ooC. are shown in Figs. 14 to 17, and are 
representative of both the gauge and hot- 
shank portions of the test bar. A fine pre 
cipitate has developed in alloys 3125 (Fig. 
14) and 3124 (Fig. 15), while in alloy 3123 
a marked increase in amount and particle 
size is evidenced from a comparison in Figs. 
12 and 16, The two types of precipitate 
found in the normalized structure of 2867 



* By the term recrystallization temperature, 
the authors infer the temperature above which 



the effects of a given strain do not accumulate 
but are removed concomitantly. 



CHARLES R. AUSTIN AND CARL H. SAMANS 



223 



are much more clearly revealed (Fig. 17) 
after prolonged testing at 600 C. 

With the exception of alloy 2867, little 
difference, except for the usual coarsening 



in a considerable increase in the amount of 
the fine precipitate. 

In raising the test temperature to 8ooC. 
the changes in microstructure of some of 







FIGS 



EFFECTS OF NORMALIZING TREATMENT OF 15 MINUTES IN HYDROGEN AT 9 soC. 

UPON MICROSTRUCTTJRES ETCHED IN MIXED ACID SOLUTION. X 75- 

Fig. io. Alloy 3125 (3.79 P er cent Fe 2 Ti). 
Fig. ii. Alloy 3124 (11.90 per cent Fe 2 Ti). 
Fig. 12. Alloy 3123 (15.62 per cent Fe 2 Ti). 
Fig. 13. Alloy 2867 (cobalt-nickel-Fe 2 Ti). 



of precipitate, was observed in the struc 
tures obtained after testing at 7ooC. as 
compared with those at 6ooC., and no 
differences could be recorded between the 
gauge cross sections and those of the hot 
shanks. However, with alloy 2867 the 
increase in test temperature from 600 to 
7ooC. resulted not only in an increase in 
the size of the spheroidal particles but also 



the alloys as well as the differences observed 
as a result of stress or plastic deformation 
were much more marked.* 

The alloy containing the smallest amount 
of ferrotitanium addition (3.8 per cent in 
3125) showed a structure similar to those 



* These facts are brought out more clearly 
in a direct study of several areas under the 
microscope. 



224 



STUDY OF SOME ALLOYS OP COBALT, IRON, AND TITANIUM 



found after testing at the lower tempera- shown. Inasmuch as an examination of the 
tures, whether the section was taken from transition section disclosed no sharp de- 



the gauge length of the test specimen marcation between the structures of the 



(Fig. 1 8) or from an unstressed sample gauge and hot shank, it would seem that 




FIGS. 14-17. EFFECTS OF TESTING AT 6ooC. UPON MICROSTRUCTURES. ETCHED IN MIXED ACID 

SOLUTION. X 750. 

Fig. 14. Alloy 3125, transverse gauge section. 
Fig. 15. Alloy 3124, transverse gauge section. 
Fig. 1 6. Alloy 3123, transverse gauge section. 
Fig. 17. Alloy 2867, transverse gauge section. 



(Fig. 19). Alloy 3124 did show certain 
differences, however. The gauge structure, 
shown in Fig. 20, is seen to have large 
particles of precipitate outlining the grain 
boundaries. A comparison of this and the 
unstressed material (Fig. 21) indicates that 
the particles are somewhat smaller in the 
latter. The structure of the hot shank was 
intermediate between the two and is not 



the plastic deformation in the gauge length 
could either have accelerated the agglomer 
ation of the precipitate or produced a more 
rapid rate of precipitation on a limited 
number of grain-boundary particles. 

The structures of 3123, the gauge of 
which is illustrated in Fig. 22 and the hot 
shank in Fig. 23, indicate that here plastic 
deformation apparently has tended to 



CHARLES R. AUSTIN AND CARL H. SAMANS 



225 



mmM^s&sqm 










FIGS. 18-23, EFFECTS OF TESTING AT 8ooC, UPON MICROSTR.UCTURES. ETCHED IN MIXED ACID 

SOLUTION. X 750. 

Fig. 18. Alloy 3125, transverse gauge section. Fig. 21. Alloy 3124, heated unstressed. 
Fig. 19. Alloy 3125, heated unstressed. Fig. 22. Alloy 3123, transverse hot-shank section. 

Fig. 20. Alloy 3124, transverse gauge section. Fig. 23. Alloy 3123, transverse gauge section. 



226 



STUDY OF SOME ALLOYS OF COBALT, IRON, AND TITANIUM 



increase the dispersion rather than the 
agglomeration of the particles. However, 
it will be noted that in this specimen the 
precipitation seems to be general rather 
than concentrated at grain boundaries as 
in alloy 3124. Consequently, if the plastic 
strain served to increase the number of 
nuclei as well as the rate of precipitation, 
the structure would give an impression of a 
decreased degree of dispersion even though 
a greater amount of precipitation actually 
had occurred. This may be what actually 
happened. 

A specimen of 2867 held at 800 C. un 
stressed for a time comparable with those 
of the other specimens was similar in struc 
ture to 3124 except for the presence of 
the numerous large particles, probably 
NiaTi, which have been commented upon 
previously. 

SURFACE STABILITY 

The four alloys differed noticeably in the 
character and degree of their surface sta 
bility during the test. The three cobalt- 
ferrotitanium alloys all showed indications 
of intergranular attack although their re 
sistance seemed to increase with increase 
in the amount of ferro titanium. The scale 
formed was not very adherent. On the 
other hand, the cobalt-nickel-ferro titanium 
alloy expanded somewhat during the test 
apparently because of a marked penetra 
tion of oxygen or nitrogen at the surface, 
and the alloy was free from selective grain- 
boundary attack. 

CONCLUSIONS 

1. The alloys investigated are susceptible 
to precipitation-hardening. 

2. 6ooC. is a favorable aging tempera 
ture regardless of the solution treatment 
used. 

3. With the 1150 and i3ooC. solution 
treatments some precipitation-hardening 
occurs even at temperatures apparently in 
the solid solution range. 



4. Tensile deformation rates at high 
temperature are affected by the oxidation 
resistance and by the amount and degree 
of dispersion of the precipitates. 

5. Both 600 and 700 C. appear to be 
within the strain-hardening temperature 
range of the two richer alloys for the 
stresses studied. However, the alloy con 
taining the smallest amount of ferro- 
titanium strain-hardened only at the lower 
temperature. 

6. NiaTi seems more prone to precipitate 
discontinuously at the grain boundaries 
while Fe 2 Ti occurs more frequently as a 
general precipitate within the grains. 

7. From microexamination it is con 
cluded that at the higher test temperatures 
(700 and 8ooC.) stress may have accel 
erated the precipitation reactions. 

ACKNOWLEDGMENT 

Mr. Robert M. Allen, a graduate student 
in metallurgy, assisted greatly in certain 
parts of the metallogra phic work, and in 
the heat-treatments employed to establish 
the quasi-binary diagram. 

REFERENCES 

1. C. R. Austin: High-temperature Properties of 

Nickel-cobalt Iron-base Age-hardening Alloys, I. 
Trans. Amer. Soc. Metals (1936) 24 (2), 451. 

2. J. Lamort: Titaneissnlegierungen. Ferrum 

(1013/14) n, 225-234- 

3. W. Koester and W. Geller: Das System Eisen- 

kobalt-titan. Archiv Eisenhuttenwesen (1934-35) 
8, 471-472. 

4. W. Jellinghaus: Kristallistruktur der Verbindung 

FesTi. Ztsch. anorg. alleg. Chem. (1936) 2.27 (i), 
62-64. 

5. H. Witte and H. J. Wallbaum: Thermische and 

Rpentgenographische Untersuchung in System 
Eisen-titan. Ztsch. Metallkunde (1938) 30, 100. 

6. R. H. Harrington: The Role of Strain in Precipita- * 

tion Reactions in Alloys. Amer. Soc. Metals 
Symposium on Precipitation-hardening, 1939. 

7. C. R. Austin and H. D. Nickol: Comparison of the 

Tensile Deformation Characteristics of Alloys 
at Elevated Temperature. Jnl, Iron and Steel 
Inst. (1938) 137, 177-221. 

DISCUSSION 

(C. S. Barrett presiding) 

J. TUCKER MACKENZIE,* Kearny, N. J. In 
the microstructures of the materials held 2000 
to 3000 hr. at 600 to 8ooC., there appears to 
be a similarity between some of the precipitates 

* U. S. Steel Corporation, Research Laboratory, 
Kearny, N. J. 



DISCUSSION 



227 



obtained by the authors and various kinds of 
nitride precipitates, principally iron nitride. 
Inasmuch as nitrogen is quite soluble in iron 
at these temperatures, it would seem possible 
that sufficient nitrogen, from the air atmosphere 
used, diffused into the alloys and was pre 
cipitated as iron nitride particles. 

In the 1936 Proceedings of the International 
Acetylene Congress, Seferian 8 and Portevin and 
Seferian 9 show photomicrographs of various 
types of iron nitride precipitates. Their work, 
in addition to recent experience of the dis 
cusser, would suggest that the precipitates 
exhibited in the photomicrographs cited might 
possibly be a form of iron nitride. In any event, 
this interesting point should be investigated, 
in order to ascertain, if possible, the true 
identity of the precipitates. 

As a tool for such an investigation, a modifi- 



8 D. Seferian: Influence de 1 Azote sur la Soudabilite 
des Aciers (Influence of Nitrogen on Weldability of 
Steel). Proc. Int. Acetylene Congress (1936). 

9 A. Portevin and D. Seferian: Etude de la Forma 
tion des Nitrures Metalliques dans la Spudure des 
Aciers (Formation of Metallic Nitrides in Welding 
of Steels). Proc. Int. Acetylene Congress (1936). 



cation of Fry s etching reagent might be 
suggested. This reagent, consisting of anhy 
drous cupric chloride, hydrochloric acid and 
ethyl alcohol, has been found by the writer to 
be the most satisfactory reagent for etching 
iron nitrides. 

C. R. AUSTIN AND C. H. SAMANS (authors 
reply). Mr. MacKenzie s comments are very 
interesting, especially since effects such as those 
mentioned by him have often been overlooked 
in the past. In the commercial development of 
these alloys the possibility of nitrogen penetra 
tion certainly must be carefully considered. 
We, however, have been unable to show that 
the needlelike structure in our specimens is 
due to the precipitation of nitrides. The modi 
fied Fry s reagent suggested by Mr. MacKenzie 
seems to etch the structure exactly as did the 
mixed acid reagent used for the illustrations in 
the paper; hence we are inclined to believe that 
the structure is due to a coarse precipitate of 
an iron-titanium compound rather than to 
nitrides formed during the long exposure to air 
at the elevated test temperature. 



Effect of Cold-work upon Electrical Conductivity of 
Copper Alloys 

BY D. K. CRAMPTON,* H. L. BuRGHOFF,t MEMBERS A.I.M.E., ANI> J. T. STACYJ 

(New York Meeting, February 1941) 



THE effect of cold-working upon elec 
trical conductivity of copper and of copper 
alloys appears not to be generally known in 
detail. Although several papers on the 
subject have been presented, showing vari 
able effects according to alloy composition, 
it is often assumed that the 2 to 3 per cent 
lowering of the conductivity of ordinary 
tough-pitch copper by severe cold-working 
is typical for copper alloys. Because of this 
situation it was decided to carry out the 
present work upon several types of copper, 
copper-zinc alloys over the commercial 
range of brasses, other binary alloys, and 
numerous age-hardening alloys in the 
quenched, and quenched and aged condi 
tions. The large number of alloys investi 
gated made it impossible to study every 
one in the detail it may have merited. Inas 
much as several interesting points that 
could not be satisfactorily explained came 
to light, further work is contemplated by 
the authors. 

It is usually considered that the normal 
effect of cold-work on ductile metals and 
unordered alloys is to decrease conductivity 
slightly. In alloys that are in the ordered 
state, as annealed or otherwise heat- 
treated, cold-- working tends to disorder the 
structure, resulting in very marked lower 
ing of conductivity. The random state of 
atomic arrangement produces lower con- 



Manuscript received at the office of the Institute 
Nov. 6, 1940. Issued as T.P. 1290 in METALS TECH 
NOLOGY, February 1941. 

* Director of Research, Chase Brass and Copper 
Co., Inc., Waterbury, Conn. 

t Research Metallurgist, Chase Brass and Copper 
Co., Inc. 

J Research Assistant, Chase Brass and Copper Co., 
Inc. 



ductivity than the ordered state for the 
same composition. 1 

PREVIOUS WORK 

Previous work in the authors laboratory 
has shown the electrical conductivity of 
several copper alloys to be decreased by 
greater amounts than might have been 
expected, in view of the general knowledge 
on the subject, when cold- worked. 

Guillet and Ballay 2 investigated the 
influence of cold-work upon the resistivity 
of several metals and alloys, including 
copper and some copper alloys. The in 
crease of resistivity due to severe working 
w is less than 4 per cent for all pure metals 
tested, while the increases for the alloys 
were generally higher. In the copper-zinc 
alloys, they found that the increase of 
resistance due to drawing increased with 
zinc content, and reported a 21.7 per cent 
increase of resistance for an alloy contain 
ing 67.9 per cent Cu as severely worked. 

Masima and Sachs 3 studied the change 
in electrical resistance of three brass crys 
tals when stretched in the tensile test, 
obtaining increases in resistance of the 
order of 3 per cent for 30 to 40 per cent 
reduction of area of the specimens. The 
resistance decreased again when the 
cold-worked crystals were subsequently 
annealed. 

Bardenheuer and Schmidt 4 found that 
the conductivity of copper was decreased 
by 3 to 4 per cent when reduced 50 per cent 
by working. The conductivity of 72-28 



228 



1 References are at the end of the paper. 



D. K. CRAMPTON, H. L. BURGHOFF AND J. T. STACY 



22Q 



brass, rolled 75 per cent, increased 7 per 
cent when annealed 4 hr. at 2ooC. and 
increased 17 per cent when similarly an 
nealed at 225C. 

Tammann and Dreyer 5 investigated the 
recovery of electrical resistance of copper 
and several binary copper alloys as well as 
other materials. Their materials were cold- 
worked 96 to 98 per cent and then heated at 
increasing temperatures. The decrease in 
resistance from the cold-worked to the 
completely recrystallized state was 20.5 per 
cent for an alloy containing 72 per cent 
Cu and 28 per cent Zn. 

TEST MATERIAL AND PROCEDURE 

The alloys for the investigation included 
four types of copper, several binary alloys 
of copper with zinc, tin, aluminum, nickel 
and silicon, respectively, as well as a num 
ber of binary and ternary age-hardening 
copper-base alloys. Electrolytic copper, 
zinc and nickel were used in the preparation 
of the alloys and other addition elements 
were used in as high a degree of purity as 
was commercially available, in order to 
have a minimum amount of impurities in 
the finished alloys. All of the alloys used 
were analyzed spectrographically for im 
purities. With few exceptions, total im 
purities were less than o.oi per cent. About 
0.02 and 0.04 per cent Fe were present in 
the two copper-silicon alloys of higher 
silicon content. In the copper-beryllium 
alloy, 0.08 per cent Al and 0.04 per cent Fe 
were present. About 0.04 per cent Al and 
Fe each were present in the copper-zir 
conium alloy. The copper-cobalt-phos 
phorus and copper-titanium-silicon alloys 
contained 0.03 and 0.02 per cent Fe, 
respectively. 

With the exception of the four types of 
copper, which were taken from commercial 
sections, the alloys were cast with labora 
tory high-frequency melting equipment in 
the form of rods iK in. in diameter. These 
were processed to o.204-in. dia., on which 
size they were annealed or otherwise heat- 



treated in an automatically controlled 
electric furnace. A drawing series was then 
made for each item, yielding material as 
drawn with reductions of area of o per cent 
and approximately 37, 60 and 84 per cent. 
The alloys in any one group were processed 
simultaneously in order to avoid possible 
variations in test results arising from slight 
departures from nominally identical 
treatments. 

Alloys of the compositions shown in 
Tables i and 2 were subjected to conduc 
tivity tests and, in some cases, tensile 
tests. Only one sample was available per 
item for electrical conductivity determina 
tion, but check tests were made by reversal 
of the specimens in the bridge. For all 
except the high-conductivity coppers, this 
meant that the actual potential determina 
tions were made in two different parts of 
each specimen. The tensile tests were made 
upon the conductivity specimens, two and 
sometimes three items being tested per 
condition. 

The electrical conductivity tests were 
made upon a Hoopes conductivity bridge, 
accurate to a reading of 0.15 per cent. The 
tensile tests were made upon a new and 
recently calibrated hydraulic testing ma 
chine, using the appropriate range in each 
case. This machine was accurate to well 
within 0.5 per cent over the testing range. 

The annealing of all material was for 
periods of i hr. unless otherwise indicated. 
Specimens were air-cooled after annealing, 
except a few items, which were quenched 
and are so indicated in the tables and 
discussions. 

DISCUSSION OF RESULTS 
Copper 

The effect of cold-drawing upon the con 
ductivity of electrolytic tough-pitch copper, 
high-conductivity Lake copper, oxygen- 
free high-conductivity copper, and phos- 
phorized copper is shown in Fig. i. For 
phosphorized copper (0.029 per cent P) 
the decrease in conductivity due to working 



230 EFFECT OF COLD-WORK UPON ELECTRICAL CONDUCTIVITY OF COPPER ALLOYS 



is almost negligible, being less than i per 
cent as drawn 84 per cent. Oxygen-free 
high-conductivity and Lake coppers show 
an almost identical decrease in conduc- 



regular increase of tensile strength is shown 
in Fig. 3. The change of conductivity due 

TABLE i. Composition and Electrical Con 
ductivity in the Annealed Condition of 
















Work-hardening Alloys Used 


"^^ 


^^F 








Alloy 
Xo. 

^ 


Composition 


Temper 
ature of 
Anneal, 
Deg. C. 


Elec 
trical 
Con 
ductiv 
ity An 
nealed, 
Per 
Cent 
I.A.C.S. 


^o- 


~__ 


o 


-0 












/O^^x tj*-r~ss**t xi f j4j-^t~ A f->fs x^ tj-j~ 



M 



FIG. i. CHANGE OF CONDUCTIVITY OF FOUR 

TYPES OF COPPER AS ANNEALED AND DRAWN. 

Open circles, electrolytic tough pitch. 
Filled circles, Lake tough pitch. 
Open squares, oxygen-free high-conductivity. 
Filled squares, phosphorized- 

tivity, which is somewhat greater than for 
the phosphorized copper. Electrolytic 
tough-pitch copper shows the greatest 
effect of these four materials, the con 
ductivity having decreased about 3.5 per 
cent as drawn 84 per cent. The effect of 
cold-working upon the conductivity of 
these coppers is small, and in general 
agreement with values previously reported 
for copper. There are very definite differ 
ences among the materials, as the present 
results indicate, for reasons which are not 
apparent. 

The percentage change in conductivity 
referred to for these materials and all 
others reported here is referred to the 
original conductivity for each material 
regardless of its actual value. The change 
is not to be construed as an arithmetical 
difference in terms of I.A.C.S. 

Copper-zinc 

The results for the various tests made 
upon copper-zinc alloys for zinc contents 
ranging up to approximately 40 per cent 
are shown in considerable detail in Figs. 2 
to 7, inclusive. It was found (Fig. 2) that 
the conductivity of each of these alloys 
decreased regularly as percentage reduction 
by drawing increased. The corresponding 



COPPERS 



I 


Electrolytic Tough 


525 


102. I 




Pitch 






2 


Oxygen Free High 
Conductivity 


525 


101 ,4 


3 


Phosphorus Deoxidized 


525 


83-9 




o . 029 % P 






4 


Tough Pitch Lake 


525 


100.5 




0.015 % Ag 







ALLOYS 











Elec 


Alloy 
No. 


Cu. 

Per 
Cent 


Alloying 
Elements, 
Per Cent 


Temper 
ature of 
Anneal, 
Deg. C. 


trical 
Con 
ductiv 
ity An 
nealed, 










I.A.C.S. 


5i 


89.65 


10.35 Zn 


525 


43-8 


52 


84-70 


15 30 Zn 


525 


36.9 


53 


79.6i 


20.39 Zn 


525 


32 9 


54 


74-18 


25.82 Zn 


525 


30. I 


55 


69.63 


30.37 Zn 


525 


28 5 


56 


64.63 


35-37 Zn 


525 


27. i 


57 


60. 17 


39.83 Zn 


525 


28.8 


7 


74-44 


25,56 Zn 


525 


30.0 








700 


30 I 


41 


70.00 


30.00 Zn 


900 Q 


27-9 








900 


28.4 








900 FC a 


28.5 


39 


59-78 


40.22 Zn 


425 Q tt 


29-3 








525 Q a 


29-t 








700 Q a 


30-3 


9 


97-40 


2 . 44 Sn 


525 


32. I 


II 


91-15 


8.68 Sn 


525 


13-5 


12 


97.52 


2 . 49 Al 


52S 


24-3 


13 


94-97 


4-99 Al 


525 


17.8 


14 


92.53 


7-50 Al 


S2S 


15-9 








900 Q 


16.1 








1010 Q a 


15-9 


IS 


89.48 


10. 08 Ni, 0.49 Mn 


700 


ii . i 


16 


79-53 


20.02 Ni, 0.46 Mn 


700 


6-5 


17 


75-82 


23-38 Ni, 0.48 Mn 


700 


5-7 


18 


70.24 


29. ii Ni, 0.50 Mn 


700 


4-7 


19 


99.01 


0.95 Si 


600 


18.4 


20 


97-95 


I - 95 Si 


600 


10.5 


21 


96.79 


2 . 99 Si 


600 


8.0 



a Q, water-quenched; FC, furnace-cooled. Where 
no qualifying symbol is given, alloys were air-cooled. 

to drawing increased with zinc content to 
2 5 to 35 per cent Zn, after which, the effect 



D. K. CRAMPTON, H. L. BURGHOFF AND J. T. STACY 



231 



for the higher zinc alpha alloys became 
less (Fig. 40)- As drawn 37 per cent, 61 per 
cent and 84 per cent, the maximum effects 
were at about 25.8 per cent Zn (approxi 
mately Cu 3 Zn), 30 per cent Zn and 35 per 
cent Zn, respectively. The alloy containing 



is at least qualitatively related to the effect 
upon tensile strength as seen by comparison 
of Fig. 4a, showing conductivity change as 
a function of zinc content, and Fig. 46, 
which shows tensile-strength change (Ib. 
per sq. in.) as a function of zinc con- 



.Composition and Initial Electrical Conductivity of Age-hardening Alloys Used 



Alloy No. 


Composition, Per Cent 


Temperature 
of Quench, 
Deg. C. 


Aging Treatment 


Electrical 
Conductiv 
ity as Heat- 
treated, Per 
Cent I.A.C.S. 


Cu 


Other Elements 


Time, Hr. 


Tempera- 
ure, Deg. C. 


22 


98.78 


1.24 Cd 


600 
600 


4 


250 


88.5 
88. 2 


23 


97.10 


2.65 Fe 


1030 
1030 


8 


650 


21.0 
40.9 


26 


98. 12 


i . 89 Ti 


900 
900 


4 


500 


8.9 

18.2 


27 


98.35 


1-53 Be 


800 
800 


24 


350 


21.2 

31-6 


28 


99.21 


0.53 Zr 


800 
800 


8 


450 


61.5 
78.3 


30 


98.93 


0.76 Fe, 0.24 P 


900 
900 


4 


450 


36.5 

69.1 


32 


98.79 


0.87 Co, 0.16 P 


900 
900 


4 


450 


34-1 

51-7 


35 


96.33 


2.92 Ni, 0.69 Si 


900 
900 


4 


450 


17-7 
32.2 


36 


99-27 


0.43 Ti, 0.22 Si 


900 
900 


8 


650 


32.8 
55 5 


25 


97 03 


2.85 Co 


950 
950 
950 
950 


i 

2 

8 


600 
600 
600 


17.2 
37-5 
38.6 
39-3 














30 . 3 


29 


98.49 


1. 17 Ni, 0.24 P 


800 
800 


i 


400 


46.9 








800 


3 


400 


53-8 








800 


8 


400 


58.7 








800 


I 


500 


59-0 








800 
800 


3 

8 


500 
500 


61.3 
62.9 








950 

950 
950 
950 


I 
8 


400 
400 
400 


30.7 
48.1 
54.7 
58.8 








950 


I 


500 


59-1 








950 
950 


8 


500 
500 


61 . 2 
63-1 



approximately 40 per cent Zn showed 
appreciable changes in conductivity when 
drawn, which appeared to vary with respect 
to the adjacent alpha alloys. This effect is 
probably due to difference in strain-harden 
ing, as will be seen in the later consideration 
of strain-hardening characteristics of the 
group of alloys. 
The effect of working upon conductivity 



tent. These two groups of curves nave 
similar characteristics, as there is little 
change between o and 10 per cent Zn, 
followed by more abrupt changes between 
10 and 20 per cent Zn, and relatively less 
change thereafter. 

Just as the conductivity and strengthen 
ing effects change sharply between 10 and 
20 per cent Zn, it is to be remembered that 



232 EFFECT OF COLD-WORK UPON ELECTRICAL CONDUCTIVITY OF COPPER ALLOYS 



other properties of copper-zinc alloys, such 
as color, corrosion resistance with particu 
lar regard to the dezincincation type of 
corrosion, and liability to stress corrosion 






FIG. 2. CHANGE OF CONDUCTIVITY OF COPPER- 
ZINC ALLOYS AS ANNEALED AND DRAWN. 

failure or season cracking, also change 
rapidly in this range of composition. 

As annealed at 525C. before drawing, 
the maximum tensile strength of the alpha 
copper-zinc alloys occurred at 25.8 per cent 
Zn, a feature that is clearly shown in Fig. 3. 
This is partly due to a grain-size effect, for 
grain size for this anneal increased with zinc 
content. The 40 per cent Zn alloy was 
naturally strongest of all at this point, 
however, because of the presence of beta. 
The relative tensile strengths of the alpha 
alloys are not essentially changed when 
drawn 37 per cent, 61 per cent and 84 per 
cent, and the 25.8 per cent Zn alloy always 
has maximum strength. The definitely 
lower rate of hardening of the 40 per cent 
Zn alloy in drawing as compared with the 
high-zinc alpha alloys is evident, for the 
40 per cent Zn alloy, which initially was 
strongest of all, constantly lost rank and 
finally had the same strength as the 35 per 
cent Zn alloy at 84 per cent reduction. 

Figs. 46 and 4^ show that, although the 
actual increase in tensile strength (Ib. per 
sq. in.) increased smoothly to a maximum 
value as zinc content increased and then 



decreased slightly in the alpha range, the 
percentage increase in strength as a func 
tion of zinc content was quite different. 
The curves in the latter case proceed to 
maxima at about 15 to 20 per cent Zn, 
then to minima at 25.8 per cent Zn, to 
maxima again at 30 to 35 per cent Zn, 
followed by definite decrease again for 
higher zinc contents, particularly 40 per 
cent Zn. These effects show the variation 
in strain-hardening according to zinc con 
tent and appear to be quite real. 

The strength and strain-hardening char 
acteristics of the 25.8 per cent Zn alloy 
are distinctive and suggest it to be a 
critical composition. It corresponds, of 
course, to a 3:1 copper-zinc atomic ratio. 
In addition, the large decrease of con 
ductivity of these copper-zinc alloys 
naturally suggests the possibility of the 
existence of ordered structure. The present 
results neither prove nor disprove this 
possibility. Conductivity changes due to 
work are indeed profound, as would be 
expected if ordering were present. Although 
there is no definite evidence for CugZn 
ordered structure, the maximum effect 
upon conductivity for the group when 
drawn 37 per cent is found at this com 
position. The fact that the maximum effect 
appears to move to alloys of higher zinc 
content for greater degrees of working 
may be directly associated with the greater 
rate of strain-hardening for the higher- 
zinc alpha alloys. 

Although the existence of ordered struc 
ture has been established in the beta 
range of copper-zinc alloys, 6 there has 
been little suggestion of an ordered struc 
ture corresponding to CusZn. Based upon 
a study of 70-30 brass, Wilson 7 has sug 
gested the possibility of such an ordered 
structure at high pressure. 

Experiments in addition to those made 
on material annealed at 525C, and drawn 
were also made. The effect of annealing 
temperature on the CusZn composition 
was thus studied by making an additional 



D. K. CRAMPTON, H. L. BTJRGHOFF AND J. T. STACY 



233 



anneal at 700 C., followed by the same 
drawing series as used for the other items. 
No essential difference in effect of working 
upon conductivity was observed for this 



containing approximately 40 per cent Zn 
was available for this purpose. This alloy 
was annealed at 425 and 525C., temper 
atures just below and just above the J3-J3 




FIG. 3. TENSILE STRENGTH or COPPER-ZINC ALLOYS AS ANNEALED AND DRAWN. 



anneal. To determine the existence of a 
possible critical point of ordering as well 
as effect of manner of cooling from the 
annealing temperature, a study was made 
by annealing 70-30 Cu-Zn for i hr. at 
900 C., followed by quenching, air-cooling 
and furnace-cooling. The electrical con 
ductivity for all these three treatments was 
essentially the same as for the 525C. 
anneal. There were, however, very slight 
differences, probably due to cooling strains, 
for the quenched material had slightly 
lower conductivity than the others. This 
small effect disappeared in the drawn 
tempers. The effect of drawing upon con 
ductivity was naturally similar to that for 
material annealed at 525C. Assuming 
that the quenching operation actually 
retained the structure formed at high 
temperature, it would appear, therefore, 
that the critical point of ordering, if 
any, does not lie below 9ooC., a tempera 
ture that is not far from the melting point 
of the material. 

The effect of beta in alpha -f- beta alloys 
was investigated in further detail than 
has already been discussed. A second alloy 



transformation, and 700 C., quenched 
and drawn. The temperature for the /3-/3 
transformation is the critical temperature 
of ordering, 6 The conductivity for all 
treatments was considerably affected by 
drawing (Fig. 5). The 35 per cent draw 
showed the same effect for both 425 and 
525C. anneals, but for 60 and 84 per cent 
draws the change in conductivity for the 
material annealed at 525C. became in 
creasingly greater than for that annealed 
at 425C. As annealed at 7ooC. and 
drawn, the lowering of conductivity was 
definitely and increasingly greater for the 
35, 60, and 84 per cent draws. Conductiv 
ity in this last case was lowered 27.5 per 
cent when drawn 84 per cent. The harden 
ing effects due to drawing this alloy in all 
conditions were similar, so that the effects 
upon conductivity appeared to be due to 
the increasing amount of beta and not to 
any differences in strain-hardening. Prob 
ably there were no differences in the char 
acteristics of the beta as quenched from 
the various temperatures. Sykes and Wilk 
inson, 6 as well as others, have concluded 



234 EFFECT OF COLD-WORK UPON ELECTRICAL CONDUCTIVITY OF COPPER ALLOYS 



that the /3-/3 transformation in brass is 
not appreciably retarded by quenching. 

The effect of annealing the copper-zinc 
wires as drawn 84 per cent was investigated 



above. Fig. 6 shows that the conductivity 
for any alloy in this annealing series 
increased continuously but not lineally, as 
the annealing temperature increased. The 




^ 



FIG. 4 CHANGE IN CONDUCTIVITY AND TENSILE STRENGTH OF COPPER-ZINC ALLOYS AS ANNEALED 

AND DRAWN. 

Open circles, 37 per cent draw. 
Filled circles, 61 per cent draw. 
Square, 84 per cent draw. 



from iooC. through the recrystallization 
range and finally at 525C. This material 
was from the drawing series discussed 



effect was noticed even for the anneal at 
iooC., for all alloys except the one con 
taining 25.8 per cent Zn. The slope of 



D. K. CRAMPTON, H. L. BURGHOIT AND J. T. STACY 



235 



these curves, which show the increase of 
conductivity as a function of annealing 
temperature, increases as annealing tem 
perature first increases, then decreases 
somewhat during the so-called relief- 
annealing range for most alloys, increases 
sharply in the recrystallization range, and 
then changes but little to the 525C. 
anneal. The 40 per cent Zn alloy was an 
exception in the annealing series. The 
conductivity of this alloy rose to a maxi 
mum value at 280 and 300 C., a value 
higher than for the original 525C, anneal, 
and decreased again for anneals at 425 
and 525C. This effect was checked by 
results for other material of similar compo 
sition and appeared to be real, although 
the reason for the effect has not yet been 
determined. For all the copper-zinc alloys 
there is considerable increase in conductiv 
ity in the relief-annealing range, so that 
it is possible to have improved combina 
tions of strength and conductivity as 
compared with the as-drawn condition. 

There appeared to be nothing critical at 
25.8 per cent Zn in this anneal except 
extremely vaguely for the lowest tempera 
tures, 100, 160 and i8oC, where less 
30 



With regard to changes of tensile 
strength in the annealing series (Fig. 7), 
there appears to be a recovery or softening 
effect for 10 and 15 per cent Zn as anneal- 



k 




-JO 



FIG. 5. CHANGE IN CONDUCTIVITY OF 
COPPER-ZINC ALLOY CONTAINING 39.8 PER CENT 
ZINC (ALPHA + BETA STRUCTURE) AS ANNEALED 
AND QUENCHED AT 425C., 525C., AND 7OOC. 
AND DRAWN. 

ing temperature increases to 180 and 
2ooC., respectively, after w r hich there is 




FIG. 6. CHANGES IN CONDUCTIVITY or COPPER-ZINC ALLOYS, PREVIOUSLY DRAWN 84 PER CENT, 

AS ANNEALED FOR ONE HOUR AT TEMPERATURES INDICATED. 

change was wrought in the conduc- a slight hardening effect before recrystal- 
tivity of this alloy than for neighboring lization and softening, although the tensile 
compositions. strength does not quite attain the value 



236 EFFECT OF COLD-WORK UPON ELECTRICAL CONDUCTIVITY OF COPPER ALLOYS 



as drawn. This apparent superposition of 
hardening upon recovery effects may con 
ceivably be due to incipient nucleation, in 
which submicroscopic nuclei serve to pro- 



6500, 7000, 1 1000 and 5500 Ib. per sq. in. 
for the alloys containing approximately 20, 
25, 30, 35, and 40 per cent Zn, respectively, 
and are more or less in line with the losses 




FIG. 7. TENSILE STRENGTH OP COPPER-ZINC ALLOYS, PREVIOUSLY DRAWN 84 PER CENT, AS 
ANNEALED FOR ONE HOUR AT TEMPERATURES INDICATED. 



duce a precipitation-hardening effect. For 
the higher zinc contents, there are definite 




FIG. 8. CHANGE IN CONDUCTIVITY OF 
COPPER-TIN, COPPER-NICKEL AND COPPER- 
SILICON ALLOYS AS ANNEALED AND DRAWN. 

increases in strength prior to softening and 
recrystallization, which amount to 1000, 



of conductivity suffered by these alloys 
when drawn 84 per cent. There is no 
quantitative relation between increase 
of conductivity and increase of tensile 
strength in this low-temperature annealing 
range. 

Copper-tin 

Two copper-tin alloys were tested with 
results as shown in Fig. 8a. The conductiv 
ity of the alloy containing 2.44 per cent Sn 
decreased 2.2 per cent as drawn 84 per 
cent, an effect of the same magnitude as 
shown by the coppers. The alloy containing 
8.68 per cent Sn showed greater effect, the 
conductivity decreasing 7.5 per cent as 
drawn only 74 per cent. 

Copper-nickel 

Four copper-nickel alloys ranging from 
10 to 29 per cent Ni were tested and 
results are shown in Fig. 86. Each alloy 
contained approximately 0.5 per cent Mn 
as deoxidizer. Although these alloys showed 
consistent trends with regard to conductiv 
ity decrease due to drawing, the conductiv- 



D. K. CRAMPTON, H. L. BURGHOFF AND J. T. STACY 



237 



ity changes for these high-resistance 
copper-nickel alloys are of about the same 
order as the accuracy of the measurement 
of conductivity. The decrease of conductiv 
ity is not great in any case and appears to 
be entirely a normal work-hardening effect. 
There is no evidence of ordered structure 
for the alloy containing 23.4 per cent Ni, 
although the copper and nickel in this 
alloy are in such ratio as to permit Cu 3 Ni 
arrangement. 

Copper-silicon 

Copper-silicon alloys containing 0.95, 
1.95 and 2.99 per cent Si showed conduc 
tivity decreases of i per cent, 5 per cent 
and 9 per cent, respectively, as drawn 
84 per cent. The results are shown in 
Fig. 8c. The trends in the group were 
consistent and conductivity decreased as 
percentage of silicon increased and as 
reduction by drawing increased. 

Copper-aluminum 

Fig. 9 shows the results obtained on three 
copper-aluminum alloys containing 2.49, 
4.99 and 7.50 per cent Al. Drawing was 
found to have very great effect on con 
ductivity, the results being comparable to 
those for the copper-zinc system. The 
alloy containing 7.50 per cent Al decreased 
in conductivity by 27.5 per cent when 
drawn 75 per cent, an amount suggestive 
of some factor other than simple lattice 
distortion. This alloy is a composition 
within the range considered suitable for 
ordering in accordance with the theory of 
Easthope as quoted by Nix and Shockley. 1 
As the possibility of some kind of ordering 
in these alloys was thus suggested, the 7.5 
per cent Al alloy was additionally annealed 
at both 900 and ioioC, quenched and 
drawn as before. The conductivity results 
were essentially the same as for the mate 
rial annealed at S25C. and drawn. The 
temperature ioioC. is just below the 
melting point for this composition, al 



though theoretically in the alpha + beta 
field, while Qoo c C. is definitely in the 
alpha field. These results indicate that, if 




\ 



\ 



2.4SA/ 



20 so 60 eo 



r 

iv* -/<? 

I 

I -a- 

L 



-30 



FIG. 9. CHANGE IN CONDUCTIVITY OF 
COPPER-ALUMINUM ALLOYS AS ANNEALED AND 

DRAWN. 

ordering is present, there is no critical 
point of ordering below ioioC. 

Age-hardening Alloys 

A number of age-hardening alloy systems 
were investigated, a typical composition 
for each system being selected. In most 
cases one quenching and one aging treat 
ment were used, which were suited to the 
particular composition. 

It was observed that the curves for the 
change in electrical conductivity as a 
function of reduction of area by drawing 
after a solution treatment followed three 
general patterns. In the first group there 
was a more or less regular decrease in 
conductivity with increasing draw. The 
second group was characterized by in 
creased conductivity for lower degrees of 
reduction, followed by a decrease for the 
higher reductions. In the third group, the 
lighter draws resulted in reduction of con 
ductivity, while more severe reductions 
lowered conductivity less or even caused an 
actual increase. 



238 EFFECT OF COLD-WORK UPON ELECTRICAL CONDUCTIVITY OF COPPER ALLOYS 



The first case, for which conductivity 
decreases regularly with degree of cold- 
working, is an example of the normal effect 
shown above for copper and its alpha 




FIG. 10. CHANGE IN CONDUCTIVITY OF 
COPPER-CADMIUM, COPPER-BERYLLIUM AND COP 
PER-NICKEL-SILICON ALLOYS AS HEAT-TREATED 
AND DRAWN. 

alloys and requires no further comment at 
this point. In the second case, the initial 
rise of conductivity may be explained by 
a precipitation from the supersaturated 
solution induced during the cold-working. 
The ensuing decrease of conductivity is 
due to the predominance of the normal 
effect of cold-working. The third case is in 
reality similar to the second and the rise 
in conductivity for the higher degrees of 
reduction may be due to a belated precipi 
tation effect in working. There is, of course, 
no assurance that this rise in conductivity 
would be continued by further working. 
It is probable that for very extreme degrees 
of working, the normal effect of working 
would again predominate. 



In the group characterized by a regular 
decrease in conductivity with increased 
working of quenched materials are Cu-Cd, 
Cu-Be, Cu-Ni-Si, and Cu-Co (Figs. 10 and 
n). Copper-cadmium alloys decreased reg 
ularly with drawing for both as-quenched 
and as-quenched and aged conditions, the 
effect being slightly greater than for 
copper. There were substantially no differ 
ences in properties of the alloy for either 
heat-treatment, which apparently is to be 
expected from the nature of the system. 8 

The conductivity of the quenched cop 
per-beryllium alloy was decreased by 1 1 per 
cent when drawn 84 per cent. When the 
alloy was aged for 24 hr. at 35oC. it was 
rendered too hard for drawing, and hence 
no results are available for such treatment. 

The conductivity of the copper-nickel- 
silicon alloy for both heat-treatments de 
creased regularly when drawn, showing 
maximum changes of 6.8 and 12.7 per cent 
for the quenched and the aged conditions, 
respectively, as drawn 84 per cent. 

The characteristic of the second group 
is that the conductivity of quenched 
material first rises and then decreases with 
increasing severity of draw. Copper-iron, 
copper-iron-phosphorus and copper-titani- 
um-silicon fell in this group (Fig. 12). 
The maximum increase in conductivity 
with the iron alloy was found to be i .9 per 
cent at 35 per cent reduction, that for 
Cu-Fe-P, 1.4 per cent at 59 per cent 
reduction, and that for the Cu-Ti-Si 
alloy 3.7 per cent at 37 per cent reduction. 
With 84 per cent reduction by drawing, the 
reduction in conductivity for these three 
alloys was found to be 9.5, 0.8 and 6.1 per 
cent, respectively. For all three materials 
as aged, the reduction in conductivity 
with increasing draw was fairly regular, the 
maximum reductions with 85 per cent 
reduction being 2.4, 11.4 and 2.5 per cent 
for the three alloys, respectively. 

In the group characterized by an initial 
reduction in conductivity when drawn 
from the quenched condition followed by 



D. K. CRAMPTON, H. L. BITRGHOFE AND J. T. STACY 



239 



an upturn in the curve with more se 
vere reduction, we ind copper-zirconium, 
copper-titanium, copper-cobalt-phosphorus 
and copper-nickel-phosphorus. The magni- 





40 



60 30 



*s 

I-" 

S-vtf 



-& 



JsJ-Zf 

*s 

c^ 

^ -J^? 



FIG. ii. CHANGE IN CONDUCTIVITY OF 

COPPER-COBALT ALLOY AS HEAT-TREATED AND 
DRAWN. 

tude of these effects is obvious from Figs. 13 
and 14. The curves for the material drawn 
after aging appear to be quite normal and 
appear to need no particular comment. 

The copper-cobalt and copper-nickel- 
phosphorus systems were studied in greater 
detail with regard to aging conditions. In 
Fig. ii the results for the copper-cobalt 
alloy are shown. Attention is called to the 
magnitude of the effect when drawing this 
alloy from the quenched condition. A 
reduction in conductivity of 33.7 per cent 
was found for a drawing reduction of 84 per 
cent. This is the largest proportionate 
change found in any of the alloys studied. 
Such an unusual change is not to be ex 
pected in such a dilute alloy where the 
possibility of ordering according to the 
ordinary conception is remote. 

Conductivity of the aged material de 
creased regularly although to a much 
smaller extent than for the quenched 



material when drawn, and the actual 
decrease bore some relation to the aging 
time, for the effect upon conductivity 
became generally less as aging time in- 




-J0 



FIG. 12. CHANGE IN CONDUCTIVITY OF 
COPPER-IRON, COPPER-IRON-PHOSPHORUS AND 
COPPER-TITANIUM-SILICON ALLOYS AS HEAT- 
TREATED AND DRAWN. 

creased. The conductivity effect would 
therefore seem to be dependent upon 
degree of precipitation, as well as actual 
degree of strain-hardening. 

The copper-nickel-phosphorus alloy was 
quenched from 800 and 950 C. and aged 
at 400 and 5ooC. for periods of i, 3 and 
8 hr. As mentioned before, the conductiv 
ity of the material as quenched from 800 C. 
decreased very slightly when drawn by 
amounts up to 60 per cent, then increased 
by 2.5 per cent when drawn 84 per cent. 



240 EFFECT OF COLD-WORK UPON ELECTRICAL CONDUCTIVITY OF COPPER ALLOYS 



When aged at 4ooC. after quenching from 
8ooC., the conductivity decreased regu 
larly in drawing, the decrease becoming 
less as the aging time increased. When 



-s 



J.63T/ 



\ 




\J -J 



^\ 


H 


\ 


1*02 


rO 




V- 

90(. 


^Q^ 


&\ 










<*z\ 
V 


\ 



2O 40 SO 60 



-/s 



FIG. 13. CHANGE IN CONDUCTIVITY OP 
COPPER-TITANIUM, COPPER-ZIRCONIUM AND COP 
PER-COBALT-PHOSPHORUS ALLOYS AS HEAT- 
TREATED AND DRAWN. 

aged at 500 C. and drawn, similar effects 
were obtained and the conductivity de 
crease due to drawing became less as the 
aging time increased. This group aged at 
5ooC. showed somewhat less effect due 
to drawing than did the group aged at 
4ooC. The results for the alloy as quenched 
from 95oC, aged and drawn, were similar 
to those discussed above although the 
trends were not quite so well defined. These 
effects may all be associated with the 
degree of precipitation, for the effect of 
drawing upon conductivity appears to 
become less as the degree of precipitation 
increases. 



SUMMARY AND CONCLUSIONS 
The effect of cold-working upon the 
electrical conductivity of several types of 
copper and of numerous copper alloys has 




FIG. 14. CHANGE IN CONDUCTIVITY OF 
COPPER-NICKEL-PHOSPHORUS ALLOY AS HEAT- 
TREATED AND DRAWN. 

been determined for reductions of area by 
drawing up to about 84 per cent. 

The lowering of conductivity of copper 
by such working was found to be in general 
agreement with previously reported results 
and was less than 4 per cent for reduction 
of area up to 84 per cent. Slight differences 
were found among the several commercial 
types of copper tested, and may be due to 
presence or absence of oxygen, silver or 
phosphorus. 

For copper alloys of the solid solution or 
work-hardening type, the results ranged 
from changes comparable with those for 
copper to far greater changes, according to 
composition and degree of working. In 
general, it was found that the lowering of 
conductivity increased with alloy content 
and with degree of draw. Very marked 
changes occurred in copper-zinc and 
copper-aluminum alloys. 

In the alpha range of copper-zinc alloys, 
the effect of cold-work upon conductivity 
increased greatly for alloys containing more 
than 10 per cent Zn, reaching a maximum 
in the range 25 to 35 per cent Zn, depending 
upon degree of working, and thereafter 
decreasing. An alloy containing 25.8 per 
cent Zn appeared to be critical and the 



DISCUSSION 



241 



large conductivity effects over a wide range 
of zinc content may possibly have been due 
to ordered structure corresponding to 
CusZn, a composition which this alloy very 
closely approximated. 

The copper-zinc alloys as drawn 84 per 
cent showed continual recovery of con 
ductivity when annealed for periods of i 
hr. at temperatures ranging from iooC. 
and upward through the recrystallization 
range, the most rapid changes occurring 
during recrystallization. 

In the alpha + beta range of copper-zinc 
alloys, the lowering of conductivity by 
working was also great and became more 
pronounced as the proportion of beta in 
creased. The existence of ordered structure 
corresponding to CuZn has previously 
been demonstrated and probably this 
accounts for much of the effect found in 
the alpha + beta range. 

The conductivity of several age-harden 
ing alloys in the quenched or solution- 
treated condition was found to be increased 
for some degrees of drawing, apparently 
because of precipitation from the super 
saturated solution in drawing, although the 
normal lowering effect predominated for 
other degrees of draw. Other alloys did not 
yield such variable results. In the age- 
hardened state, the decrease of conduc 
tivity was roughly proportional to the 
degree of draw. This is in accordance with 
the results obtained upon simple work- 
hardening alloys, which were at no time 
supersaturated. The degree of saturation 
of the aged alloys was certainly much less 
than for the quenched alloys. The magni 
tude of the effect was also dependent upon 
the degree of precipitation, for the change 
became less as aging time increased and 
also as aging temperature increased. 

ACKNOWLEDGMENTS 

The authors wish to make due acknowl 
edgment to Messrs. E. L. Smith and P. A. 
Leichtle for the necessary chemical and 
spectrographic analyses, and to Mr. H. F. 



Schissler for assistance in the preparation 
of the alloys. Thanks are also due to the 
Chase Brass and Copper Co. for permission 
to publish the test results. 

REFERENCES 

1. F. C. Nix and W. Shockley: Reviews of Modern 

Physics (1938) 10, i. 

2. L. Guillet and M. Ballay: Compt. rend. (1923) 

176, 1800. 

3. M. Masima and G. Sachs: Ztsch. Physik (1928) 51, 

321. 

4. R. Bardenheuer and H. Schmidt: Mitt. Kaiser- 

Wilhelm-Inst. Eisenforsch. (1928) 10, 193. 

5. G. Tammann and K. L. Dreyer: Ann. Physik 

(1933) 16, 657. 

6. C. Sykes and H. Wilkinson: Jnl. Inst. Metals 

(1937) 61, 223. 

7- T. C. Wilson: Phys. Rev. (1939) 56, 598. 
8. E. A. Owen and L. Pickup: Proc. Roy. Soc. (1933) 

I39-A, 526. 

DISCUSSION 

(E, E. Schumacher presiding) 

W. C. ELLIS,* New York, N. Y. The 
authors have presented in one paper a collection 
of valuable data pertaining to the effect of cold- 
work on the conductivity of copper and some 
important copper alloys. In addition to being 
of engineering interest, the data are also 
important in supplying fundamental informa 
tion that should be useful in relationship to the 
theory of metallic conduction. In this connec 
tion the large effect of cold-work in decreasing 
the conductivity of certain copper alloys, in 
particular those of zinc and aluminum, in 
comparison with the effect on copper, is 
notable. 

It is well established that in metallic con 
ductors electrical resistivity basically results 
from a scattering of the conduction electrons 
and that this scattering increases with the 
degree of departure from periodicity in the 
metallic lattice. One could explain the greater 
effect of cold- work in solid solution alloys as 
resulting from a greater disturbance in the 
perfection of the lattice periodicity occasioned 
by cold-work for the alloys than for the pure 
metal. Such an explanation does not carry far, 
for the reason that the greater departure from 
perfection in the alloys is still obscure. Other 
factors that may affect the resistivity change 
by cold-work are: structural changes accom 
panying deformation (grain fragmentation) 
and distribution changes in the two kinds of 



* Bell Telephone Laboratories. 



242 EFFECT OF COLD-WORK UPON ELECTRICAL CONDUCTIVITY OF COPPER ALLOYS 



atoms, grossly, or over volumes very much 
smaller than the metallographic grain. In this 
connection Mott has pointed out that a random 
distribution of atoms may not be the one 
having the highest resistance. If small nuclei 
are present of about the magnitude of an 
electron wave (10 A.) scattering may be more 
effective than for larger or smaller nuclei. 

It is stated in the paper that in the copper- 
zinc alloys, which were the ones most exten 
sively studied, the effect of working on 
conductivity is at least qualitatively related to 
the effect of working on tensile strength. Can it 
be assumed that the same internal process 
that is, the mechanism of plastic deformation 
is responsible for the change in each of these 
properties? If so, an understanding of the cause 
for conductivity changes might go far toward 
illuminating the obscurity existing in the de 
pendence of strength properties on plastic 
deformation. The authors are to be con 
gratulated upon having provided data that 
may stimulate consideration of the plastic 
deformation process from a viewpoint electri 
cal-conductivity changes that perhaps has not 
received the emphasis it merits. 

S. E. MADDIGAN,* Waterbury, Conn. On 
page 236 the authors have suggested that in 
cipient nucleation caused the small but definite 
hardening effect at temperatures below the 
visible recrystallization range. Such a concept 
was previously suggested by the writer 9 to 
explain a small anomalous drop in conductivity 
previous to visible recrystallization in a long 
time annealing study. In this work the anoma 
lous conductivity effect appeared greatest in 
the lower ranges of reduction, whereas the 
hardening effect was much greater in the high 
reductions. This at first seems to be paradoxical 
if both effects are explained as due to incipient 
nucleation. Actually, however, there is no 
discrepancy. 

Because of the much higher nucleation rate 
(as evidenced by the subsequent smaller an 
nealed grain size), any effect due to incipient 
nucleation should increase with increasing 
degree of reduction. This is found to be true 
for the hardening effect. However, during the 
formation of incipient nuclei a recovery process 
is occurring in the matrix. This apparently has 

* Research Physicist, Chase Brass and Copper Co. 
S. E. Maddigan and A. I. Blank: Trans. A.I.M.E. 
(1940) 137, 170. 



little effect on the tensile strength, but is known 
to cause large increases in conductivity. 
Probably for high reductions this large re 
covery change masks out the much smaller 
change postulated as due to incipient nuclea 
tion. The nucleation effect might, however, 
be visible even for large reductions if observa 
tions were made at sufficiently frequent time 
intervals. Incidentally, the incipient nuclei as 
discussed later might be ordering nuclei rather 
than recrystallization nuclei. 10 

The authors have suggested that the mass 
of their results on copper-zinc alloys indicate 
an ordering effect at CuaZn. They have wisely 
not attempted to offer this work as conclusive 
proof of such ordering. Certain points in the 
paper require further consideration in their 
relation to the ordering effect. 

In Fig. 40 a minimum in the percentage of 
tensile strength change is found at the Cu 3 Zn 
composition. This agrees with the suggested 
ordering, as ordering is sometimes accompanied 
by a hardening effect (p. 56 of ref. i). Such an 
ordering might also be responsible for the 
hardening effect in the recovery range (Fig. 7). 
In the previous work of the writer, 9 a double- 
stepped change in conductivity was found 
during annealing as shown in Fig. 15. If two 
separate processes of nucleation and growth 
are assumed, a similar curve results, as shown 
in the figure. In the calculations the equation 
F = A(i e~ Kt } suggested by Johnson and 
Mehl 11 for processes of nucleation and growth 
has been used, with the values A 0.07 and 
0.15; K = 20 and 0.02. It is known that con 
ductivity is a function of ordering 10 and if it is 
assumed that conductivity is also a function 
of the recrystallization process such a curve is 
readily explained. Since the ordering process 
requires an exchange of atom positions while 
the recrystallization process requires only re- 
orientation and some slight motion of atoms, 
the first stage should be recrystallization, the 
second stage ordering. This would agree with 
the frequent observation that complete recov 
ery of conductivity occurs in many cases at 
relatively low annealing temperatures. 

Since both recrystallization and ordering are 
linked with diffusion, it is difficult to under 
stand why such great differences should exist 



10 J. Bardeen: Jnl. Applied Physics (1940) n. 88- 

11 W. A. Johnson and R. P. Mehl: Trans. A.I.M E 
(1939) 135. 416. 



DISCUSSION 



243 



in the K values used here. It might be sug 
gested that recrystallization is associated not 
with diffusion but with the vibrational energy 
of the lattice. In this case the recrystallization 



paper. After the so-called recovery stage, a 
straight-line relation also exists. Changes in 
conductivity with ordering are usually non 
linear in relation to percentage reduction. 1 



Calculoted curves K20 
Sum of calculated curvet 
Experimental annealing 



a 0.02 



10 01 



Time in hours 



100 



10 OO 



FIG. 15. EXPERIMENTAL ANNEALING CURVE COMPARED TO CALCULATED NITCLEATION CURVE. 



temperatures for various metals 12 would be 
related to the Debye characteristic tempera 
tures. 13 However, a comparison shows no rela 
tionship between the magnitudes of these two 
temperatures, whereas for the few cases given 
in the literature the energy of activation of 
diffusion increases in the same order as the 
recrystallization temperature. 

It is also difficult to believe that the second 
stage of the conductivity change is not associ 
ated with recrystallization when the first 
visible new grains appear just at the beginning 
of this second steep rise in conductivity. On 
the other hand, it can hardly be considered 
that the first stage shown can be due to order 
ing. It should further be mentioned that the 
conductivity change in each stage is greater 
than the 2 to 3 per cent usually observed in 
other metals and alloys. 

Fig. 2 shows a straight-line relation between 
percentage of reduction and conductivity 
change. The writer s work gave the results 
shown in Fig. 16. After complete annealing, a 
straight-line relation exists as in the present 

12 R. S. Archer: Metals Handbook (1939) 200. 
Amer. Soc. Metals. 

13 F. Seitz: Modern Theory of Solids. New York. 
McGraw-Hill Book Co. 



This makes it doubtful whether either stage of 
the conductivity change can be associated with 
ordering. 

In Fig. 6, for the iooC. anneal the 75 per 
cent Cu-25 per cent Zn alloy shows less con 
ductivity change than any of the others. 
Assuming ordering, this can be explained only 
by the formation of ordering nuclei even at 
this low temperature. 10 At such a temperature 
the diffusion coefficient is extremely low 14 and 
it seems doubtful that an observable inter 
change of atom types could occur in short-time 
anneals. It would therefore seem that an 
explanation of this point would require some 
process other than ordering. 

The authors have mentioned the results of 
T. C. Wilson for 70-30 brass under high pres 
sures. It is not clear how an ordering effect 
could appear at high pressures if it already 
exists at normal pressures with a critical tem 
perature apparently above 900 C. This sug 
gests that the annealing phenomena observed 
at normal pressures are not due to ordering. 

In Fig. ii an extremely large conductivity 
change is observed for the copper-cobalt alloy. 



"P. N. Rhines and R. F. Mehl: Trans. A.I.M.E. 
(1938) 128, 185. 



244 EFFECT OF COLD-WORK UPON ELECTRICAL CONDUCTIVITY OF COPPER ALLOYS 



As pointed out by the authors, this can hardly 
be related to ordering. It may possibly be 
associated with a precipitation induced by 
working. One of the chief reasons for suggesting 



H.L. BURGHOFF (author s reply). Mr. Ellis 
suggestion that the mechanism of plastic de 
formation may be responsible for the change in 
both strength and conductivity of materials is 






Per cent Reduction 
40 



60 



FIG. 1 6, EFFECT OF DEGREE OF REDUCTION ON THE CONDUCTIVITY CHANGE FOR RELIEF ANNEALING 

AND COMPLETE ANNEALING. 



ordering in the copper-zinc alloy is the large 
conductivity change effected by cold-work. In 
contradiction to this, we now have a still 
greater conductivity change in an alloy of 
which the explanation almost certainly lies 
elsewhere. In addition, the high conductivity 
changes in tungsten and molybdenum are well 
known. These are pure metals and hence do not 
involve ordering. 

Of the various points discussed here, some 
seem to support the postulate of an ordered 
structure of the composition Cu 3 Zn, while other 
points seem to support the opposite conclusion. 
A neutron transmission study such as that of 
Nix and Dunning 16 would be of great value in 
reaching a conclusion on this subject. 



la Nix, Beyer and Dunning: Phys, Rev. (1940) 58, 
1031. 



interesting. It leaves a large field open for 
experimentation. Of course, as everyone knows, 
the actual mechanism of plastic deformation 
forms an elusive subject and is still being 
studied by numerous investigators. 

Our work was of necessity, covering so many 
alloys, rather general in nature with regard to 
many of the materials. I do believe that a more 
careful study of the strain-hardening character 
istics of all of them would be fruitful. 

Dr. Maddigan has discussed various phases 
of the paper. If an alloy with ordered structure 
is deformed by cold-working, the ordering is 
reduced or possibly destroyed, owing to a 
relative displacement of atoms, which need not 
be great. The reordering during annealing 
would not require great atom movement and 
could coincide with recovery and recrystalliza- 
tion effects. 



DISCUSSION 



245 



The lack of change in conductivity of the 
75-25 Cu-Zn alloy when annealed at iooC. is 
suggested to be due to formation of ordering 
nuclei. The small effects observed in material 
annealed at this low temperature, if real 
and capable of being differentiated, would 
seem to call for no active process in this 
alloy. 



The large conductivity change reported for 
the quenched and drawn copper-cobalt alloy 
represents data for an isolated case and is 
worthy of further work to see if the results 
can be duplicated. The explanation of the large 
change does not lie in any brittleness of the 
alloy, as probably it does for the relatively 
brittle metal tungsten. 



Low-temperature Oxidation of Single Crystals of Copper 

BY BENJAMIN LUSTMAN,* JUNIOR MEMBER, AND ROBERT F. MEHL,f MEMBER A.I.M.E. 

(Ne-w York Meeting, February 1941) 



THE study of the high- temperature oxida 
tion of pure metals, intensively pursued 
experimentally since the pioneer work of 
Pilling and Bedworth 1 and supplemented 
by the recent theoretical work of Wagner 2 
is now well advanced. The important field 
of low-temperature oxidation i.e., oxida 
tion extending only through the temper 
color range of thicknesses with the forma 
tion of thin oxide layers in the range from 
o to 1000-2000 Angstroms has not, how 
ever, reached so satisfactory a state, owing 
largely to the experimental difficulties that 
attend the formation and study of thin 
films; most of the work has been performed 
on abraded specimens whose surface 
irregularities approximated the thickness 
of the films formed. It has been pointed out 
that the rates of oxidation of different 
crystal faces are different and that the 
difference may be related to the orientation 
relationships that obtain between the metal 
crystal and the superimposed oxide crystal ; 
it has also been suggested that this orienta 
tion dependence should manifest itself in 
rates of oxidation and perhaps also of 
corrosion in aggregates bearing preferred 
orientations. 3 - 4 An attempt to measure the 
rates of oxidation of different crystal faces 
of alpha iron demonstrated this difference 
in rate, 4 but the dependence of rate on 
orientation was not simple; the method of 



Submitted by Benjamin Lustman to the faculty of 
the Carnegie Institute of Technology in partial ful 
fillment of the degree of Doctor of Science, June 1940. 
Manuscript received at the office of the Institute 
Nov. 7, 1940. Issued as T.P. 1317 in METALS TECH 
NOLOGY, April 1041. 

* Department of Metallurgy, Carnegie Institute 
of Technology, Pittsburgh, Pa. 

t Department of Metallurgy, Carnegie Institute of 
Technology, Pittsburgh, Pa. 

1 References are at the end of the paper. 



measuring thickness, consisting of a color 
comparison with barium stearate multi- 
films of known thickness, was not wholly 
satisfactory. 

For these reasons it was felt that a study 
of the rate of oxidation of single crystals 
prepared with surfaces as smooth as 
possible, using a sensitive method giving 
true film thickness, should be profitable. 
Copper was chosen because of the ease of 
reduction of the primary oxide film and 
because the orientation relationships be 
tween metal and oxide and the composition 
of the oxide films formed at low tempera 
tures are known; for thin films formed at 
low temperatures, Cu 2 O alone occurs. 22 " 24 

EXPERIMENTAL PROCEDURE 

Various methods of surface preparation 
were attempted in order to obtain a surface 
that would oxidize uniformly and repro- 
ducibly. Electrolytic polishing by the 
method of Jacquet 5 and electrolytic and 
mechanical polishing followed by annealing 
in vacuum and in purified hydrogen at 
various temperatures were tested. The 
procedure finally adopted consisted in a 
fine metallographic polish followed by an 
anneal in purified hydrogen at temperatures 
of 900 to ioooC. for 12 hr. or more. 
The surface so produced was very bright 
and smooth, with no trace of specular 
etching (though some grain-boundary etch 
ing was observed); indeed, fine scratches 
produced by the final polish were entirely 
removed by the high-temperature anneal. 
After this treatment, the specimens were 
cooled in hydrogen to room temperature 
and placed in the oxidation furnace, where 



246 



BENJAMIN LUSTMAN AND ROBERT P. MEHL 



247 



the air-formed film (never more than 10 to 
20 A. thick, as measured by the method 
described below) was removed by hydrogen 
reduction at 3SoC. The specimen was 



Laue back-reflection method after cutting, 
polishing, and annealing; these are shown 
in Fig. 2 plotted in a unit stereographic 
triangle. 




2000 



4000 



6000 



8000 



TIME IN SECONDS 

FIG. i. REPRODUCIBILITY OF OXIDATION CURVES. 



then cooled to the oxidation temperature 
and dried air was admitted at a vigorous 
rate of flow. The air and hydrogen purifica 
tion trains are shown in Fig. 3. 

The reproducibility of oxidation curves 
obtained from surfaces prepared in this 
manner is shown in Fig. i. The curves for 
crystal 3 and crystal o were obtained from 
the same specimens after repolishing and 
reannealing. The oxidation curves for 
specimen i were obtained from two differ 
ent crystals of the same orientation. 

Oxygen-free high-conductivity copper 
was used and single crystals were prepared 
by the Bridgman method. The orientations 
of their surfaces were determined by the 



The apparatus used, including the puri 
fication train for the hydrogen and air, is 
shown schematically in Fig. 3. The speci 
mens, of cylindrical shape, fitted snugly 
into a hole drilled into a block of copper 
around which the furnace windings were 
placed; the purpose of this arrangement 
was to provide the specimen with a source 
of heat in order to prevent a drift in the 
temperature of the specimen as the change 
from a hydrogen to an air atmosphere was 
made. The temperature of the crystals 
during oxidation was controlled within 
limits of HC. 

The incident light source was plane- 
polarized and reflected from the con- 



248 LOW-TEMPERATURE OXIDATION OF SINGLE CRYSTALS OF COPPER 



tinuously oxidizing metal surface. On 
reflection from a metal surface, plane- 
polarized light is generally transformed to 
elliptically polarized light, the amount of 




100 HO 

FIG. 2. ORIENTATION or COPPER CRYSTALS. 

whose ellipticity depends on the refractive 
index of the metal and on the angle of 
incidence. The effect of a transparent film 
on the metal is to change the amount and 
nature of the ellipticity, which in the 
present case was measured by means of a 
quarter-wave plate and analyzing Nicol 
prism, with the addition of a half-shade 
system to increase the sensitivity of 
settings. By comparison of theoretical 
and actual ellipticity changes, both the 
refractive index and the thickness of the 
film could be obtained. Relations between 
the change in ellipticity and the film thick 
ness and index of refraction were derived 
in a previous publication, by Leberknight 
and Lustman. 6 The absolute accuracy of 
film thickness determination is controlled 
principally (indeed, almost exclusively) by 
the precision of measurement of the film 
index of refraction; knowledge of the latter 
to 10 or 20 per cent permits a determination 
of film thickness with an accuracy of i to 
5 per cent.* 



* Tronstad" applied a similar experimental 
method to the study of films on metals; his 
interpretation of the polarimetric data, how 
ever, is valid only for very thin films, less than 
50 A. thick. 



This method possesses several advan 
tages over procedures previously employed 
to determine rates of film formation on 
metals. In the first place, its sensitivity of 
measurement is considerably greater than 
that of the other methods: thickness 
changes of one Angstrom unit may be 
detected. The method used by Hinshel- 
wood 7 and others, 8 - 9 involving measure 
ment of change in pressure as the metal 
oxidizes at constant volume, is capable of 
comparable sensitivity at low pressures. 
However, this sensitivity is lost at high 
pressures and is hardly capable of accurate 
measurement at pressures above o.i 
atmospheres; furthermore, changes occur 
ring within the film with changing pressure 
render the use of such a method question 
able. The present method possesses the 
further advantage of allowing continuous, 
nondestructive measurement of film thick 
ness during oxidation; this is of particular 
and obvious utility in determining the 
kinetic laws of film-forming processes. Of 
the methods commonly used, only that 
involving measurement of pressure changes 
and the temper-color method (consisting 
essentially of a comparison of the inter 
ference colors formed by the film with some 
standard) possess this same advantage. The 
applicability and disadvantages of the 
latter have been thoroughly reviewed by 
Evans; 10 its insensitiveness, inaccuracy, 
and inapplicability to films below 300 to 
400 A. in thickness are serious drawbacks. 
In accuracy, the present method is ap 
proached only by the electrometric method 
of Miley 11 and Campbell and Thomas; 12 
the latter method is capable of considerable 
accuracy and sensitivity but is destructive, 
discontinuous, and rather time-consuming. 
Thickness readings may be taken quickly 
(5 to 10 sec.), thus allowing close measure 
ment of fast processes. Finally, small metal 
specimens of almost any shape may be used 
and their surface preparation can be closely 
controlled; the gravimetric method as used 
by Vernon, 13 in which the weight of the film 



BENJAMIN LUSTMAN AND ROBERT F. MEHL 



249 




COMPRESSED 
AIR 



FIG. 3. APPARATUS, INCLUDING PURIFICATION TRAIN. 
A, telescope. 
By analyzer scale. 

C, analyzing Nicol. 

D, half-shade. 

E, quarter-wave plate scale. 

F, quarter-wave plate. 

G, copper block furnace. 



H, glass windows. 

/, polarizing Nicol. 

/, collimating lens. 

K, slit. 

L, filter. 

M t lens. 

N, mercury vapor lamp. 



O, activated alumina. 
P, platinum catalyst. 
Q, tank hydrogen. 
JR., barium oxide. 
5, calcium chloride. 
T, flow meter. 



i 

o 



130 C 

ANNEAL 

CRYSTAL TEMP. 
4 750C 

9SOC 

I 950C 

t / 750C 




2000 



4000 6000 

TIME IN SECONDS 

FIG. 4. EFFECT OF ANNEALING TEMPERATURE ON OXIDATION OF COPPER, 



8000 



250 



LOW-TEMPERATURE OXIDATION OF SINGLE CRYSTALS OF COPPER 



45000 



36000 



27000 



10000 



9000 



~C? 4OOO 8000 

TtM IN SECONDS 




20 

TIME IN SECONDS 
THICKNESS IN 4 




SO fOO 500 1 000 5000 1000.0 

T/ME IN SECONDS 

FIG. 5. EFFECT OF DIFFERENT METHODS OF 
PLOTTING ON SHAPE OF OXIDATION CURVE OF 
CRYSTAL 14 AT IO5C. 



formed is determined, requires the use of 
comparatively large specimens with large 
surfaces (such as sheet), the preparation 
of which would seem to be difficult. The 
disadvantages of the method used are: (i) 
It gives no information concerning the 
composition of the films; (2) it is applicable 
only to plane reflecting surfaces and cannot 
be used to determine film thickness on dull 
abraded surfaces; (3) it loses both sensi 
tivity and accuracy with films of irregular 
thickness; (4) the determination of thick 
ness from the observed optical properties 
requires rather cumbersome and involved 
calculations. 

EXPERIMENTAL RESULTS 
Effect of Annealing Temperature 

Annealing the mechanically polished 
surface at temperatures above 95oC. 
yields a very bright, smooth surface; 
annealing at a lower temperature (75oC.) 
produces a surface showing slight specular 
etching. 

Oxidation curves at i3oC. were made 
on specimens showing both types of sur 
faces (Fig. 4), The initial rate of oxidation 
is appreciably greater for the surface 
annealed at 95oC. than for the other sur 
face, but later the rate is smaller. This 
difference, as is evident from comparison of 
Figs. 4 and i, is greater than would be 
expected from the reproducibility of the 
oxidation curves. Thus, considering the 
sensitivity of the initial rate to slight 
differences in surface, it is to be surmised 
that measurements on the severely dis 
rupted and stressed surface produced by 
abrasion should be of uncertain significance. 

Form of the Oxidation Curve 

Three principal types of equations have 
been proposed to explain the course of 
oxidation at low temperatures of metals 
forming protective films the parabolic 
law, 14 - 1 

I* = kt [i] 



BENJAMIN LUSTMAN AND ROBERT P. MEHL 



25* 



the parabolic law corrected for the velocity 
of interface reactions, 15 " 17 

kj + P = fat [2] 

and Tammann s logarithmic equation, 18 



t = 



[3] 



in which / is thickness, t time, and a, b and 
k are constants. 

To decide which if any of these laws is 
obeyed, the oxidation curve of crystal 
4 at iosC. was plotted in the three follow 
ing ways: square of thickness versus time, 
thickness versus time divided by thickness, 
and thickness versus the logarithm of time. 
The result is shown in Fig. 5 ; it is evident 
that equations i and 2 are not followed 
at any section of the curve, while, except 
for a brief initial period of less than 100 
sec., the Tammann logarithmic law applies 
throughout. 

In Figs. 6 to 10 are plotted the film thick 
ness formed on all crystals at the oxidation 
temperatures 80, 105, 118, 130 and iS5C. 
versus the logarithm of time for one 
atmosphere air pressure. While the simple 
logarithmic law is obeyed initially in all 
cases to times of 1000 to 5000 sec., devia 
tions generally occur at longer times. 
Except for crystal 9 oxidized at io5C. 
(Fig. 7) the second part of the curve 
possesses a greater slope than the initial 
part. The complete course of the curves 
may in most cases be described by an 
equation of the form 



[4] 



It should be emphasized that it is the 
second part of such curves that is observed 
by a temper-color method of assessing 
thickness and that first led Tammann to 
the formulation of the simple logarithmic 
law. 

The quantity to be chosen as expressing 
the rate constant for reaction curves obey 
ing equation 4 deserves consideration. 




V 001 HI SS 3 NX 3 1 HI 



2 S 2 



LOW-TEMPERATURE OXIDATION OF SINGLE CRYSTALS OF COPPER 



Obviously the thickness of film formed at 
any given time will depend on both quan 
tities fli and &i (excluding for the moment 
consideration of the later stages of oxida- 



tity r-> the slope of the oxidation curve 
when plotted on a logarithm of time 
abscissa scale, is more reproducible and 




so 



100 



SO O 1000 
TIME JN SECONDS 



5000 10000 




SOO 100O 
TIME IN SECONDS 

b 



5000 10000 



FIG. 7. OXIDATION OF COPPER AT io5C. ON LOGARITHMIC SCALE. 



tion) ; thus neither of the quantities ai or &i 
alone will serve completely to specify the 
rate of oxidation. However, since the quan- 



seems physically more expressive of oxida 
tion rate than the quantity a\, which is the 
intercept of the curves on the abscissa at 




SO 



100 



500 1000 

TtME IN SECONDS 



SOOO /OOOO 



Q 

o 



I 

g 




IOO 



500 fOOO 
TIME IN SECONDS 

FIG. 8. OXIDATION OP COPPEE. AT n8C. ON LOGARITHMIC SCALE. 



SOOO /OOOO 



2S4 



LOW-TEHPERATURE OXIDATION OF SINGLE CRYSTALS OF COPPER 




50 100 



soo tooo 

TIME IN SECONDS 



SO 00 10000 




too 



SOOO IOOOO 



SOO IOOO 
TIME IN SECONDS 

b 
FIG. 9. OXIDATION OF COPPER AT i3oC, ON LOGARITHMIC TIME SCALE, 



BENJAMIN LUSTMAN AND ROBERT F. MEHL 



2 5S 



to the value of 



large values of a\ accom- 

i 



zero thickness, the former was chosen as 

LU LI1C VCU.U.C UJl 7 

the rate constant. However, it should be Ol 

emphasized that because of the partial panying large values of r-; a\ varies from 

dependence of rate on the quantity ai about i sec. at 8oC. to 50 sec. at J55C. 

16 




100 



500 1000 
TfME IN SECONDS 



5000 tO 000 



: /2 



Ul 

o 

l< 



755 C 

CRYSTAL 







50 



100 



5000 10000 



500 1000 
TIME fN SECOHD& 

b 

FIG. 10. OXIDATION OF COPPER AT iss c C. ON LOGARITHMIC TIME SCALE. 

at any given time, in spite of a greater Orientation Dependence of Oxidation Rate 
value of the rate constant r-> the oxide The orientation relationships existing 
thickness attained by a given crystal may between Cu 2 and Cu after low-tempera- 
be less than that reached by a crystal ture oxidation have been thoroughly in- 
showing a smaller value of the rate con- vestigated. Mehl, McCandless and Rhines 3 
stant. In general, a\ varies proportionally found that in scale formation on copper the 



2 S 6 



LOW-TEMPERATURE OXIDATION OF SINGLE CRYSTALS OF COPPER 



Cu 2 lattice was oriented in the same way 
as the underlying copper lattice. This is 
identical with one of the orientations 
observed later by Moore, 22 who found on 



ao c 



105* C 




FIG. ii. ORIENTATION DEPENDENCE OF 

OXIDATION RATE PLOTTED IN UNIT STEREO- 
GRAPHIC TRIANGLE. 

Fine lines are contour lines of equal slope of 
logarithmic oxidation curves (1/61) and figures 
refer to magnitude of i/bi in Angstroms. 

(100) planes of copper the orientation 
relationship (in) Cu20J|(ioo) Cu and 
[no] CusOJjtioo] Cu, while he found on 
(no) and (in) planes of copper the orien 
tation relationship (100) Cu20||(ioo) Cu, 
and [100] Cu 2 O||[ioo] Cu. This confirmed 
the orientations reported somewhat ear 
lier by Yamaguchi 23 and Thiessen and 
Schiitza. 24 Thus, except for orientations 
around the (100) plane of copper, the 
orientations of the copper surfaces shown 
in Fig. 2 are those of the Cu 2 film formed 
on each surface. 

The slopes of the initial parts of the 
curves in Figs. 6 to 10 inclusive, (corre 
sponding to T- in equation 4) may be taken 



as expressing the rate of oxidation, and 
these are plotted in the appropriate posi 
tion in the unit triangle in Fig. n and 
contour lines of equal oxidation rate are 
drawn. The minimum in the rate at 8oC. 
apparent in the center of the stereographic 
triangle almost disappears at io5C., at 
which temperature the oxidation rate 
seems to be inversely proportional to dis 
tance from the (100) pole. At ii8C. the 
(in) pole apparently is a singular axis, 
the oxidation rate increasing with angular 
separation from this direction; however, 

the same trend of the quantity g- is noted 

as at io5C. i.e., (100) planes of copper 
oxidize faster than (no) planes, with (in) 
planes showing minimum rate. At isoC. 
the maximum in the oxidation rate shifts 
from the (100) pole to approximately the 
position occupied by crystal o (Fig. 2); 
the same dependence is shown at i55C, 

Temperature Dependence of Oxidation Rate 

Dunn 26 applied the Arrhenius equation, 
which has been shown to be applicable to 
all diffusional processes, 43 to the explana 
tion of the temperature variation of oxida 
tion rate. This equation states that 

W = Ae-v* T [5] 

in which W is the oxidation velocity, A the 
action constant, e the base of natural 
logarithms, Q the heat of activation, R 
the gas constant, and T absolute tempera 
ture. Thus a plot of the logarithm of 
oxidation velocity versus the reciprocal of 
absolute temperature should give a straight 
line if the oxidation rate depends on a 
simple activated diffusion through the 
oxide film. Pilling and Bed worth 1 had 
previously expressed the temperature var 
iation of scaling rate by an exponential 
equation of another form. 
In Fig. 12 the logarithm of the oxidation 

rate constant \T~) has been plotted against 
the reciprocal of the absolute temperature 



BENJAMIN LUSTMAN AND ROBERT P. MEHL 



257 



for each of the crystals studied; obviously, 
since a straight-line plot does not result, 
a growth process is occurring, which is not 
explicable by a classical temperature varia 
tion. It will be noted that a minimum in the 
oxidation rate occurs in the neighborhood 
of 105 to i55C. for each crystal face. 
It should be emphasized here that this 

minimum is peculiar to the quantity T-; 

for almost every crystal the oxide thick 
nesses formed after a given oxidation time 
increase quite normally with increase in 
reaction temperature, though again not 
according to the activation equation 
(Eq. 5). That Tammann and Koster 18 
found a discontinuity in the oxidation 
rate of copper at 205 to 22oC. and Dunn 26 
at 209 to 24iC., whereas in the present 
investigation the minimum was observed 
at 105 to issC., may perhaps be ascribed 
to differences in the character of the initial 
metal surface resulting from different modes 
of preparation. 

The oxidation-rate constants, r- for 

the latter parts of the curves of Figs. 6 to 
10 are plotted in Fig. 13; this quantity 
likewise does not follow the Arrhenius 
equation. As a measure of the rate of 
chemical attack of copper by oxygen, the 
rate of thickening of the oxide within the 
first 3 min. of oxidation was chosen; again 
the activation equation (Eq. 5) was not 
followed, as is shown in Fig. 13, 

Effect of Oxygen Pressure 

The rate of oxidation of crystal 4 was 
determined at 118 and io5C. at various 
air pressures up to one atmosphere. The 
pressures were obtained by pumping 
through suitable capillaries and constricted 
tubes, so that a vigorous rate of flow of air 
was maintained past the specimen. Fig. 14 
shows the results obtained at three values 
of the partial pressure of oxygen; a maxi 
mum in the oxidation rate with decreasing 
pressure of oxygen may be observed. This is 



better shown in Fig. 1 5 for the oxidation of 
crystal 4 at io5C. and Fig. 16 at n8C. 

The existence of a maximum rate of 
oxidation at about 0.015 atmospheres of 



2.5 



0.5 
2.5 



0.5 
2.5 



7.5 



0.5 




CRYSTAL 
3 
2 
O 
I 





CRYSTAL 
4 

6 

7 




CRYSTAL 
8 




/55t J30CIJ8C 105 C 



acre 



FIG. 12. VARIATION OF OXIDATION RATE 

(l/bi) WITH TEMPERATURE. 

oxygen may be noted for both tempera 
tures; for all pressures at these two temper 
atures the logarithmic law (Eq. 4) again is 
obeyed for the greater part of the curves. 
In Fig. 17 the oxidation-rate constant 

f^-j has been plotted against partial 

pressure of oxygen. While at higher pres 
sures crystal 4 oxidizes faster at io5C. 
than at n8C., at the lower values of 
oxygen pressure the oxidation-rate con- 



2 5 8 



LOW-TEMPERATURE OXIDATION OF SINGLE CRYSTALS OF COPPER 



slant increases normally with increase in 
temperature; undoubtedly it was for this 
reason that Wilkins and Rideal 8 (whose 
experiments were all conducted at low 



by a parabolic time law and whose rate of 
thickening increases with temperature 
according to the Arrhenius equation 
(Eq. 5) must obviously grow by a mecha- 




15 S*C I3CTC llfC 1 0$*C 



acre 



FIG. 13. VARIATION OF FINAL OXIDATION RATE (1/62) AND INITIAL RATE WITH TEMPERATURE. 



pressures) observed no minimum in 
the variation of oxidation velocity with 
temperature. 

DISCUSSION or RESULTS 

From a consideration of the complexity 
of the experimental data, it is immediately 
obvious that no simple theory based upon 
ionic and electronic diffusion such as is so 
successfully applicable to the explanation 
of scaling velocities of metals will serve to 
explain the low-temperature oxidation of 
copper. Scales that increase in thickness 



nism different from that of films that form 
according to an exponential time law and 
by a complex temperature dependence of 
growth. Therefore the following discussion 
will be restricted to a consideration of the 
similarities in the low-temperature oxida 
tion behavior of metals and of the theories 
that have been proposed. 

Since the presentation of the logarithmic 
law by Tammann and Roster in 1922, it 
has been continually attacked and ques 
tioned. Most of the objections were based 
on the method used by Tammann for its 



BENJAMIN LUSTMAN AND ROBERT F. MEHL 



259 



experimental verification the temper-color 
method objections that for the most part 
are valid. 10 However, it is not certain that 
the corrections that would have to be made 
to Tammann s experimental data would 
vitiate his conclusions as to the shape of the 
oxidation curve. It has been shown in the 
preceding paragraphs that copper oxidizes 
logarithmically in the temperature range 
80 to i55C. By the same experimental 
method a few oxidation curves also have 
been obtained for single crystals of iron 
at a temperature of 3ooC. The shape of 
these curves when plotted on a logarithmic 
time scale is shown in Fig. 18; the validity 
of the exponential law in explaining the 
oxidation of iron may be recognized. 
Similarly, Vernon, Akeroyd, and Stroud 27 
showed that below 2ooC., a temperature 
that Vernon 13 believed marked a change in 
the crystal structure of Fe 2 0s, iron oxidizes 
logarithmically. Steinheil 41 determined the 
course of oxidation of thin films of alumi 
num by the change in their transparency; 
his curves are shown replotted on a logarith 
mic time abscissa scale in Fig. 18. Vernon, 
Akeroyd, and Stroud 27 established gravi- 
metrically the applicability of the logarith 
mic law to the oxidation of zinc at 125 to 
225C. Thus Eqs. 3 and 4 have been found 
to be generally valid for the explanation 
of the low-temperature oxidation of copper, 
iron, aluminum and zinc by three very 
different experimental techniques; it would 
therefore seem correct to state tentatively 
that all metals forming protective oxides 
oxidize according to the logarithmic law 
at low temperatures. This is the first gen 
eralization in the oxidation behavior of 
metals that requires explanation. 

Several explanations have been proposed 
to account for the logarithmic law. Wag 
ner 30 considered that Tammann s law may 
be a matter of purely accidental reality and 
that it may arise from instability of the 
primary oxide layer due to the following 
causes: (i) recrystallization or sintering 
phenomena in the film; (2) transition from 



unstable to stable crystal modifications; (3) 
formation of a higher oxide at larger film 




2000 4000 6000 

TIME IN SECONDS 

a 




II f .OI78A 



O 2000 4000 6000 8000 

TIME IN SECONDS 







^*^ 


__ 


I OS*. 00 16 AT, 


^ue .0013 AT. 






\L 








2OOO 4000 000 80t 



TIME IN SECONDS 
C 

FIG. 14. OXIDATION OF COPPER AT PRESSURES 
AND TEMPERATURES INDICATED. 

thicknesses. The first possibility may be 
ruled out immediately; however valid the 
assumption of sintering might be for films 



260 



LOW-TEMPERATURE OXIDATION OF SINGLE CRYSTALS OF COPPER 



formed on. strained, abraded, polycrystal- 
line surfaces, in the present work, where 
strain-free, single- crystalline surfaces were 
used, a single-crystalline oxide must result. 



crystal 4 several orientations would result 
from such a process, since the orientation 
relationship obtaining on (100) planes leads 
to several different possibilities of match- 




V N! SS3NXO/HJ. 



Even if it is supposed that growth of the 
films occurs from a number of points on the 
metal surface, for all orientations except 
that of crystal 4 a single crystal of oxide 
must result, since the orientation of the 
oxide is the same as that of the metal. For 



ing; however, inasmuch as crystal 4 shows 
the same behavior on oxidation as all 
the other crystals, this fact would seem to 
be of little importance in determining the 
manner of oxidation. Furthermore, since 
it has been shown 3 that single- crystalline 



BENJAMIN LUSTMAN AND ROBERT F. MEHL 



261 



oxide films remain single crystalline even have shown that at Cu 2 O thicknesses less 

after they have grown into thick scales, than 400 A. the oxide consists entirely of 

recrystallization within the primary oxide Cu 2 0, while in the range 400 to 800 A. 

film is obviously excluded. Possibilities 2 traces of CuO may occur; only above 




o 



o 

z 
o 



W SSZNXOtHJ. 



and 3 may likewise be ruled out on the basis 
of electron-diffraction studies, which indi 
cate that the films formed at low tempera 
tures on copper are composed of cubic 
Cu20 predominantly at all oxidation times 
and temperatures. Cruzan and Miley 42 



800 A. thickness of film definite amounts of 
CuO always appear and then only in very 
small amounts. In the present experiments 
the oxide thickness attained 400 A. only at 
i55C.; during the initial logarithmic part 
of the oxidation curves, it may be safely 



262 



LOW-TEMPERATURE OXIDATION OF SINGLE CRYSTALS OF COPPER 



assumed therefore that the film consists 
merely of Cu 2 0. 

If it is assumed that the ionic concentra 
tion gradient within the film is an exponen 




.1 .15 

PARTIAL PRESSURE O g /H ATMOS. 

FIG. 17. VARIATION OF OXIDATION RATE (i/bi) 

OF COPPER WITH PRESSURE. 

tial function of distance, and that the 
simple diffusion laws may be applied to 
such a concentration gradient, an equation 
of the form of the logarithmic law may be 
derived; it was thus that Tammann justi 
fied his equation and that Vernon and his 
coworkers 27 explained the oxidation be 
havior of zinc at low temperatures. How 
ever, it is doubtful whether either of these 
assumptions is valid. 19 Certainly it has been 
shown definitely 31 that if oxidation results 
from movement of ions through the oxide 
in conformity with Pick s laws, with diffu 
sion velocity independent of the concentra 
tion of anions or cations, the parabolic law 
will result. There is, however, adequate 
evidence that the diffusion constant will 
not have a fixed value throughout the film. 
Rother and Bohnke 32 have shown that 
electrical conductivity in thick oxide scales 
formed on copper varies markedly within 
the oxide as shown in Fig. 19. That a 
similar condition obtains in the thin films 
formed in the present investigation may be 
inferred from Fig. 20. The method of 
determining thickness permitted also a 
rough determination of the index of refrac 
tion of the film ; it was found that the results 
could be satisfactorily expressed by con 
stancy of the real part of the index n (with 



a value of 2) and by a variation of the 
absorption index k as shown in Fig. 20. If 
it is assumed that variation of the light 
absorption within the film is due for the 
most part to a change in the number of free 
electrons, the apparent average absorption 
will tend to trace a course with increasing 
film thickness similar to the curve of elec 
trical conductivity (which is also propor 
tional to the number of free electrons or of 
missing electron sites) within the film. 
From the parallelism of conductivity and 
diffusion rate 33 one would thus expect a 
variation of the diffusion coefficient with 
concentration; if such a variation were 
introduced into the formal diffusion laws, a 
form of the resulting oxidation equation 
might well follow which would differ from 
the parabolic law. It must, however, be 
noted that in spite of the known variation 
of conductivity (diffusion rate) with con 
centration in thick films the scaling of 
copper generally follows a parabolic equa 
tion. An explanation of the logarithmic 
law by ionic diffusion processes occurring 
within the film is thus a possibility, but at 
present only an obscure one. 

Mott 39 has recently proposed a theory 
that promises to be of importance in 
explaining the low-temperature oxidation 
of metals. His theory is founded primarily 
on the premise that in thin films the ionic 
and electronic concentration gradients 
through the film may be independent of 
each other, whereas in thick films, because 
of space charges that would be set up by 
unequal gradients, the ionic and electronic 
gradients are interdependent. Mott assumes 
that in thin films the rate of film growth is 
controlled by the rate at which electrons 
can leave the metal surface and ionize 
absorbed oxygen atoms at the oxide-gas 
interface. The probability that an electron 
that leaves a metal surface with an energy 
W (the work function) will penetrate by 
the tunneling effect a barrier of height U 
and thickness / (the film thickness) is pro 
portional to the expression 



BENJAMIN LUSTMAN AND ROBERT F. MEHL 



263 



where m is the mass of the electron and h is 
Planck s constant. The derivation of the 



change in slope at 2ooC. in the plot of log 
of oxidation rate of iron versus the recipro 
cal of oxidation temperature, which again 
was attributed to a change in the crystal 



800 



I 



IRON 





500 



200 



I i 

ALUMINUM 



I0 2 



I0 3 



I0 4 



10 2 



I0 4 



a TIME IN SECONDS 

FIG. 1 8 OXIDATION OF IRON (LUSTMAN AND MEHL) AND ALUMINUM (STEINHEIL* S ) ON LOGA 

RITHMIC TIME SCALE. 



logarithmic law from such an expression is 
obvious. It will particularly be noted that 
the factor corresponding to i/bi in Eq. 4 

depends only on the 



temperature-independent factors U and W 
the energies required for an electron to 
penetrate the oxide and to leave the metal. 
Thus, by Mott s theory, no temperature 
dependence of i/&i is predicted. 

The second generalization that may be 
proposed concerns the temperature depend 
ence of oxidation rate at low temperatures. 
It was shown above that a minimum in the 
oxidation rate of copper exists at 105 to 
i55C. Dunn 20 ascribed a similar effect to 
sintering or grain growth within the film; 
such an explanation is obviously inappli 
cable to the single-crystal oxide films 
formed here. Vernon, Akeroyd and Stroud 27 
recently observed a similar minimum in the 
oxidation rate of etched abraded and 
anodically polished zinc suafaces at tem 
peratures of 120 to 225C. This effect the 
latter authors attributed to a change in the 
crystal structure of ZnO from amorphous 
below 225C. to hexagonal above this 
temperature. Vernon 13 also reported a 



structure of Fe 2 0s from the alpha form 
above 2ooC. to the gamma form below 
this temperature. In the case of the oxide 
formed on copper, numerous exhaustive 
electron-diffraction and X-ray investiga 
tions disclosed no change in the crystal 
structure of the predominant oxide phase 
Cu2O from the lowest to the highest tem 
peratures of oxidation. From the fact that 
deviations from the expected oxidation 
behavior occur within so close a tempera 
ture range, 105 to 225C., for metals as 
dissimilar as copper, zinc, and iron, forming 
dissimilar oxides, it would seem more 
inviting to attribute these common differ 
ences to some mechanism common to the 
oxidation of each of these three metals. 
Certainly no minima should occur if the 
oxidation rate is controlled by ionic diffu 
sion through the film, in which case Eq. (5) 
should hold. Mott s theory, on the other 
hand, predicts that i/bi should have a 
constant value independent of temperature. 
It is obvious that no theory yet proposed is 
adequate to explain the temperature de 
pendence of film formation on metals. Thus 
if the data on the three metals listed are 
considered illustrative of a general behavior, 



264 



LOW-TEMPERATURE OXIDATION OP SINGLE CRYSTALS OP COPPER 



it may be hypothesized that the oxidation 
rate of metals at low temperatures changes 
discontinuously with temperature. 
It was found in the experiments on copper 




Layer Thickness 
FIG. 19. VARIATION OF FREE COPPER 
CONTENT AND CONDUCTIVITY IN Cu 2 FILM 

(R.UTHER AND BoHNKE 37 ). 

that in general the oxidation curves 
resolved themselves into two straight-line 
portions when plotted on a logarithmic 
time scale, the time law consisting of a sum 
of two exponential expressions. As is evi 
dent from Fig. 18, iron and aluminum both 
show the same type of behavior. Curves 
presented for the oxidation of zinc and of 
iron (below 2ooC.) by Vernon, Akeroyd, 
and Stroud 27 indicate that similar curves 
would have been obtained if a more sensi 
tive measuring method had been used. Thus 
the third characteristic of oxide film forma 
tion may be that the course of oxidation of 
metals at low temperatures is given not by one 
but by at least two exponential expressions. 
Again none of the theories, including 
Mott s, predicts this result. 

It has been suggested that the two 
exponential sections may arise from the 
formation of CuO at longer oxidation times. 
If this were true, the oxidation velocity 



should be smaller at greater thicknesses 
because of the low ionic mobility in CuO; 
however, it has been found that the second 
logarithmic section is always of greater 
slope than the first. Further, Cruzan and 
Miley 42 have shown that CuO is not present 
in oxide films of the thicknesses produced 
in the present investigation. 

A fourth characteristic of the low tem 
perature oxidation of metals may be that 
smooth surfaces oxidize more slowly than 
rough surfaces. Thus it was shown earlier in 
this paper that an etched copper surface 
oxidizes at a slower initial rate, but faster 
final rate than a smooth surface. A similar 
effect was noted by Vernon, Akeroyd, and 
Stroud 27 in the low-temperature oxidation 
of zinc. In the latter, the oxidation curves of 
anodically polished surfaces (measured 
over long time periods) were found to lie at 
lower thicknesses than those obtained from 
either abraded, or abraded and etched 
surfaces. From the slower final oxidation 
rate of smooth strain-free surfaces as com 
pared with that of rough or etched strain- 
free surfaces, which apparently is valid for 
the low-temperature oxidation of both 
copper and zinc, one might offer the general 
ization that the effect of manner of surface 
preparation is predominantly one of sur 
face geometry, smooth surfaces reacting 
more slowly than serrated surfaces. The 
well-known effect of low-temperature reduc 
tion of an oxide film on the subsequent 
acceleration of oxidation may be attributed 
thus to a microscopic surface roughening. 
Finally, it may be stated as a fifth gener 
alization that oxidation rate varies with 
crystal orientation in a manner not to be 
predicted in a simple way from the crystallog 
raphy of the oxides or from the oxide-metal 
orientation relationships. The rate of oxida 
tion of copper single crystals at 105, 118, 
130 and i55C. was found to vary with 
angular separation from some singular axis, 
the [100] direction at io5C., [in] at 
n8C., and the pole of crystal o at 130 
and i55C. Braun 39 found that the dec- 



BENJAMIN LUSTMAN AND ROBERT F. MEHL 



265 



trical conductivity of rock salt in the [100], 
[no], and [in] directions was in the ratio 
6:4:3, which is explicable by a mechanism 
in which conduction occurs in a cube direc- 



conductivity, Diinwald and Wagner 28 con 
nected with a state of disordering of the 
oxide involving empty places in the Cu + 
lattice and in the electron lattice (electron- 




7 Q 9 10 II 



THICKNESS IN 100 A 

FIG. 20. VARIATION OF ABSORPTION INDEX or Cu 2 FILMS WITH THICKNESS. 



tion only i.e., by a mechanism in which 
the [100] direction is a singular direction, as 
in the oxidation experiments above. How 
ever, the reason for the change in the 
singular axis with oxidation temperature is 
obscure on the basis of this explanation; 
furthermore, Joffe 35 found that the con 
ductivity of cubic crystals is isotropic. 
Gwathmey and Benton 26 found, by ob 
serving the temper colors on a copper single- 
crystal sphere oxidized at various pressures 
and temperatures, even more complicated 
orientation dependencies. Mehl and Mc- 
Candless, 4 studying the oxidation of iron, 
likewise found a complex orientation de 
pendence, which again could not be recon 
ciled with relative oxide-metal orientation. 
Because of the paucity of experimental 
data on the variation of oxidation rate of 
metals with oxygen pressure, the generality 
of the effect of pressure noted for the oxida 
tion of copper cannot be assumed. It has 
been observed that at high temperatures 
the scaling rate of copper increases with 
pressure as the seventh root of that 
quantity. 29 This, in the analogous case of 



defect places). However, Schottky and 
Waibel 36 found that Cu2O shows electron 
excess conduction below 500 C., which, 
interpreted in terms of lattice disordering, 
could mean excess copper ions in the lattice 
above the stoichiometric composition (or a 
deficit in oxygen ions) and free electrons. 
This state of disordering is similar to that 
which exists in ZnO and CdO, the con 
ductivity of which Baumbach and Wagner 37 
found to increase with decreasing pressure. 
The increase in rate of oxidation with 
decreasing pressure at high pressures in 
the present investigation is thus explicable. 
It should be noted that this mechanism 
explains only the variation of oxidation 
rate at high pressures; the decrease ob 
served at lower pressures must obviously 
depend on some other process. Whatever 
this other process might be, it cannot be 
one that would occasion a radical change 
in the manner of oxidation, for the oxidation 
curves were observed to be substantially 
logarithmic throughout the range of pres 
sures investigated. It is possible, however, 
that with decreasing pressure the solid 



266 



LOW-TEMPERATURE OXIDATION OF SINGLE CRYSTALS OF COPPER 



solution composition range of Cu 2 O be 
comes narrower; the decrease in the 
concentration gradient would tend to cause 
a decrease in oxidation rate. 

MECHANISM OF INITIAL REACTION 

The process by which oxygen molecules 
react with clean metallic surfaces to form 
the initial oxide film has recently been 
studied by Roberts. 38 On exposure of a 
clean metallic surface to oxygen, a dis 
continuous two-dimensional atomic film of 
oxygen forms, and by diffusion through this 
film the initial three-dimensional oxide 
results. On hydrogen-covered tungsten 
wire, this adsorption of oxygen was found 
to be a comparatively slow process. Con 
firmation of these conclusions was found 
in the present investigation. When air was 
admitted to the specimens after reduction 
but without a preliminary evacuation of 
the hydrogen, a period of 30 to 40 sec. was 
observed during which the specimen sur 
face showed no apparent change i.e., no 
refracting film i to 2 A. thick was formed; 
after this initial period, oxidation began 
at a maximum rate. Presumably during 
this initial period the adsorbed hydrogen 
film was being replaced by an absorbed 
oxygen film. When the hydrogen was first 
evacuated the oxidation process began 
immediately at maximum rate; most or all 
of the adsorbed hydrogen film was appar 
ently driven off. 

At the very lowest pressure investigated, 
an entirely different mechanism of initial 
oxide formation is operative. From Figs. 
20, 21 and 22, it may be seen that at 
0.0016 to 0.0013 atmospheres partial pres 
sure of oxygen the oxidation curves show 
first an increasing and then a decreasing 
rate. A similar phenomenon has been 
observed by other investigators. Wagner 
and Griinewald 29 reported an initial period 
of increasing rate in the scaling of copper 
at high temperatures and low pressures. 
Bircumshaw and Preston 21 found a marked 
"induction" period in the oxidation of 



molten tin at low pressures. This effect 
of pressure may be associated with a 
change in the stability of the oxide nuclei 
and hence with a change in the rate of 
nucleation; Johnson and Mehl 44 have 
shown that curves of the shape indicated 
arise from processes of nucleation and 
growth. Thus, in spite of the greatly differ 
ent nature of the metallic bases, and of the 
structure and rate of formation of the reac 
tion products, the effect of low pressures 
in each of these cases was to change the 
nature of the initial reaction. While at high 
pressures the initial oxide forms apparently 
by an adsorption-diffusion process, at low 
pressures a process of nucleation and 
growth is operative. 

SUMMARY 

1. A sensitive method of thickness deter 
mination in the temper-color range of 
thicknesses was applied to a continuous 
study of the rate of oxidation of single 
crystals of copper at various temperatures 
and pressures for a number of crystal 
orientations. 

2. The effect of surface preparation was 
investigated and a method giving repro 
ducible surfaces was developed. Smooth 
surfaces were found to oxidize more slowly 
than rough surfaces at long oxidation times, 
more rapidly at short oxidation times. 

3 . The oxidation curves were of the form : 



where / is time; /, thickness; e, the base of 
natural logarithms; and a x , a 2 , bi and b Z} 
constants. Thus, when plotted on a loga 
rithmic time scale, the oxidation curves 
consisted of two straight-line portions, the 
second being of greater slope. 

4. The variation in oxidation rate with 
temperature was found to show a minimum 

in the constant T~> for all crystal orienta 
tions with increasing oxidation temperature. 



DISCUSSION 



267 



5. The orientation dependence of oxida 
tion rate was found to be complex and to 
depend in general on angular distance from 
a singular direction at any one temperature 
and also to vary with temperature. The 
dependence could not be correlated with 
oxide-metal orientation. 

6. The oxidation of iron crystals at 
30oC. was also shown to obey the loga 
rithmic equation. 

7. With increasing oxygen pressure, the 
oxidation rate first increases to a maximum 
and then slowly decreases. The decrease 
was explained by an electron excess type of 
disordering in Cu^O at low temperatures, 
and the increase by an expansion of the 
solid solution range of the oxide with 
increasing oxygen pressure. 

8. The initial formation of the oxide on 
clean metal surfaces may be described as 
an adsorption-diffusion process at high 
pressures, and as a nucleation and growth 
process at low pressures. This was evidenced 
by a continuously decreasing oxidation 
rate with time at high pressures, and at low 
pressures by first an increasing, and later a 
decreasing, oxidation rate. 

9. Similarities in the dependence of 
oxidation rate on time, temperature, 
crystal orientation, and surface preparation 
of iron, zinc, aluminum and copper were 
discussed in connection with a general 
theory of the oxidation of metals at low 
temperatures. 

REFERENCES 



N. B. Pilling and R. E. Bedworth: Jnl. Inst. 

Metals (1923) 29, 529. 

C. Wagner: Ztsch. Phys. Chem. (1933) 2I-B, 25. 
R. F. Mehl, E. H. McCandless and F. N. Rhines: 

Nature (1934) *34 1009. 
R. F. Mehl and E. H. McCandless: Trans. 

A.I.M.E. (1937) 125, 53i. 
P. A. Jacquet: Nature (1935) *35, 1076. 
C. E. Leberknight and B. Lustman: Jnl. Optical 

Soc. Amer. (1939) 29, 59. 
C. N. Hinshelwood: Proc. Roy. Soc. (1922) 

I02-A, 318. 
, r W A,_. " . Rideal: Proc. Roy. Soc. 



_ . w . . __, Roy. Soc. (1930) I28-A, 407. 

U. R. Evans: Metallic Corrosion, Passivity and 

Protection, 672 ff. London, 1937. Edward 

Arnold. 
H. A. Miley: Jnl. Amer. Chem. Soc. (1937) 59 

2626. 
W. R. Campbell and U. B. Thomas: Trans. 

Electrochem. Soc. (1939) 309. 



W. H. J. Vernon: Trans. Faraday Soc. (1935) 31, 

1668. 
G. Tammann: Ztsch. anorg. allg. Chem. (120) 

in, 83. 
U. R. Evans: Trans. Electrochem. Soc. (1924) 

46, 247. 

K. Fishbeck: Ztsch. Elektrochem. (1933) 39, 318. 
K. Fishbeck, L. Neundeubel and F. Salzer: 

Ztsch. Elektrochem. (1934) 4, 5*7. 
G. Tammann and W. Koster: Ztsch. anorg. allg. 

Chem. (1922) 123, 196. 
W. Jost: Diffusion und Chemische Reaktion in 

Festen Stoffen, 27 ff. Berlin, 1937. Th. Stein- 

kopf. 

J. S. Dunn: Proc. Roy. Soc. (1926) m-A, 210. 
L. L. Bircumshaw and G. D. Preston: Phil. Mag. 

(1936) 21, 686. 

K. H. Moore: Ann. Phys. (1938) 33. 133- 
T. Yamaguchi: Proc. Phys. Math. Soc. Japan 

(1938) 20, 230. 
P. A. Thiessen and H. Schultze: Ztsch. anorg. 

allg. Chem. (1937) 233, 35. 
A. T. Gwathmey and A. F. Benton: Jnl. Chem. 

Phys. (1940) 8, 431. 

J. S. Dunn: Proc. Roy. Soc. (1926) ni-A, 203. 
W. H. J. Vernon, E. I. Akeroyd and E. G. Stroud: 

Jnl. Inst. Metals (1939) 65. 
H. Dunwald and C. Wagner: Ztsch. phys. Chem. 

(1933) 22-B, 212. 

C. Wagner and K. Grunewald: Ztsch. phys. Chem. 

(1938) 40-B, 455. 

C. Wagner: Angew. Chem. (1936) 49, 735. 

K. Heindlhofer and B. M. Larsen: Trans. Amer. 

Soc. Steel Treat. (1933) 21, 865. 
R. Ruther and H. Bohnke: Ztsch. Phys. (1933) 

81, 771. 

W. Jost: Ref. 19, 92fL 
F. Braun: Ann. Phys. (1887) 31, 855. 
A. F. Joffe: The Physics of Crystals, 85. 
W. Schottky and F. Waibel: Phys. Ztsch. (1933) 

34> 855. 
H. H. V. Baumbach and C. Wagner: Ztsch. phys. 

Chem. (1933) 22-B, 199. 

J. K.Roberts: Proc. Roy. Soc. (1935) I52-A, 477. 
N. F. Mott: Trans. Faraday Soc. (1939) 35, H75; 

(1940) 36, 472. 
L. Tronstad: Det. Kgl. Vidensk. Selsk. Skrift 

(1931) No. i t i. 

A. Steinheil: Ann, Phys. (1934) *9 455- 
C. G. Cruzan and H. A. Miley: Jnl. App. Phys. 

(1940) n, 631. 

R. F. Mehl: Trans. A.I.M.E. (1936) 122, ir. 
W, A. Johnson and R. F. Mehl: Trans. A.I.M.E. 

(1939) 135, 416. 

DISCUSSION 

(L. L. IV y man presiding) 

U. R. EVANS,* Cambridge, England. The 
authors studies of the low-temperature oxida 
tion of copper will be welcome, the more so as 
they bring out well the influence of crystallo- 
graphic direction in determining the velocity 
a factor that is shown qualitatively when 
microsections of polycrystalline copper or brass 
are heated, the different grains assuming dif 
ferent tints. I only propose to deal with one 
matter, on which the authors have quoted my 
opinion; namely, the use and limitations of 
interference colors as a guide to film thickness. 

The interference color method has the un 
doubted attraction that it is possible to obtain 
"measurements without measuring." Never- 



13. 
14. 
is- 

16. 

17. 

18. 
19. 

20. 
21. 

22. 
23- 

24. 
25. 

26. 
27. 

28. 
29. 

30. 
31. 



33- 
34- 
35- 
36. 



38- 
39. 

40. 

41. 
42. 

43- 
44. 



* Cambridge University. 



268 



LOW-TEMPERATURE OXIDATION OF SINGLE CRYSTALS OF COPPER 



theless, a very elementary insight into optical 
theory indicates that, in the form of the 
method that is sometimes employed, the 
measurements will be far from accurate. If, 
however, one cares to calibrate the color scale 
by some other method, it is possible to use 
interference colors to obtain values of reasonable 
accuracy. A favorable case is that of silver 
iodide films, which were studied by L. C. Ban 
nister and myself n years ago. 45 Here five 
orders of colors can be obtained. Measurements 
of film thickness were carried out by three 
independent methods (microgravimetric, elec- 
trometric and nephelometric). It was found 
possible to employ all three methods on one 
and the same specimen, so that there was no 
question of "matching." Since good agreement 
between the three sets of measurements was 
obtained, it was fairly certain that the numbers 
were not far from the truth. Consequently, it 
was permissible to use any one of the methods 
for calibrating the color scale, and tables were 
constructed showing the range of thicknesses 
corresponding to the different tints. Some 
colors correspond to a very small range of 
thickness, and are thus a precise measure, 
whereas others represent an extensive range of 
thickness. It was found that the thicknesses 
calculated by the ordinary mode of employing 
interference tints were in many cases quite 
wrong. 

It must be remembered that different 
methods of estimating film thickness do not 
necessarily measure the same thing. Most 
optical methods (certainly those depending on 
interference tints) measure the effective dis 
tance between two reflecting surfaces. In the 
case of interlocking between metal and oxide, 
for which there is much evidence today, a 
considerable amount of oxide might be buried 
away below the effective position of the lower 
surface, and not enter into the value obtained. 
The electrometric method gives only the oxide 
accessible to the cathodic action, which may 
not represent the whole. The gravimetric 
method gives the difference between the oxide 
present at the first and second weighings. If, 
as will usually be true, a sensible amount of 
oxide is present at the time of the first weighing, 
it may sometimes be necessary to estimate this 
by another method, and add it on, if it is 

** U. R. Evans and L. C. Bannister: Proc. Roy. 
Sec. (1929) I25-A, 370. 



desired to have a measure of the total oxide 
present after heating. 

Not many years ago, there were differences 
even in the order of magnitude between the 
thicknesses as found by different methods for 
the films corresponding to the various colors. 
This state of affairs was unsatisfactory. Today 
the values obtained by all the methods are 
found to give somewhat similar numbers, but 
discrepancies remain. These may be caused 
partly by methods that are insufficiently exact, 
but probably they are largely caused by the 
fact that different methods measure different 
things. If so, the differences between the results 
of different procedures, far from being a matter 
of dissatisfaction, add to our knowledge of 
the manner in which oxide and metal pass into 
one another. It is probable that several differ 
ent methods ought to be used, before the true 
geometry of an oxidized surface can be fully 
understood. 

W. H. J. VERNON,* Teddington, England. I 
appreciate the opportunity afforded me of con 
tributing to the discussion on this valuable 
paper, while regretting my inability at present 
to offer more than a few tentative comments 
on several of the many points raised. Primarily, 
I wish to offer my heartiest congratulations on 
the advance which this paper represents both 
in refinements of technique and in the con 
clusions to which it has led. The especial ad 
vantage of the authors method, apart from 
its applicability to small single-crystal speci 
mens, is its ability to give an almost continuous 
record of oxidation rate in the very earliest 
stages, which is just where the gravimetric 
method fails. It would be of great interest to 
see their results extended so as to fill the hiatus 
between and to correlate the two groups of 
results. On the question of methods of prepara 
tion, while I fully appreciate the fundamental 
value of the technique employed by the 
authors, I am inclined to doubt their surmise 
(p. 250) that " measurements on the . . . 
surface produced by abrasion should be of 
uncertain significance." If the abrasion is care 
fully conducted, quite a high degree of repro- 
ducibility can in fact be achieved; and, after 
all, the surface produced by abrasion cannot 
be left out of account in practical oxidation 
problems. 

* Chemical Research Laboratory, Department of 
Scientific and Industrial Research. 



DISCUSSION 



269 



The application of the logarithmic law to the 
low-temperature oxidation of single crystals 
of copper, for temperatures up to i5SC., is of 
considerable interest. From gravimetric results 
(using ordinary poly crystalline surfaces pre 
pared by abrasion) it was concluded 46 47 that 
copper obeys the parabolic law at temperatures 
as low as 7 5C- (possibly soC.). Is this overlap 
to be attributed to the different physical con 
dition of the specimens, or is it essentially a 
question of time the shorter the initial period 
over which the observations are made, the 
higher being the temperature at which the 
logarithmic relationship persists? Similarly 
with iron. Gravimetric results from ordinary 
abraded surfaces showed that the logarithmic 
law changed over to the parabolic law when a 
fairly well denned transition temperature, 
about 200C., was exceeded. 48 - 49 On the other 
hand, for single crystals and short periods of 
exposure the authors have obtained logarithmic 
curves at sooC., while the transition tempera 
ture under their conditions may obviously be 
higher still. 

Experiments with zinc 49 have suggested that 
the first-formed oxide is amorphous. Is it possi 
ble that this is generally true where the 
logarithmic law is followed? The question is 
difficult to answer experimentally, having re 
gard to the small amount of oxide formed 
during such short periods at low temperatures, 
but the authors views on this matter would be 
welcomed. A further interesting question 
relates to interference colors. It seems now to 
be fairly certain that the development of the 
usual sequence of colors is characteristically 
associated with the parabolic type of oxidation 
e.g., copper over a wide temperature range, 
iron above 2ooC. and the absence of such 
colors with the logarithmic type e.g., iron 
below 2ooC.,* zinc over a wide temperature 
range. In the light of the authors results, 
copper at low temperatures is now seen to 
follow the logarithmic type; the absence of 
interference colors, at least at ordinary room 



"W. H. J. Vernon: Jnl. Chem. Soc. (1926) 2273. 
W. H. J. Vernon: Trans, Faraday Soc. (1927) 




temperature, notwithstanding that the film 
may reach thicknesses well within the inter 
ference-color range for higher temperatures, I 
can confirm from my own early work. 47 These 
related groups of phenomena are possibly 
associated with a continuous type of film for 
the parabola and a discrete type originating 
with an initial point-to-point attack for the 
logarithmic curve; for example, microscopic 
examination of the oxide film on zinc shows 
very clearly a finely granular structure, in 
contrast with the continuous film on copper at 
the higher temperatures. The authors work 
should throw considerable light on this ques 
tion, and their conclusions will be awaited with 
interest. 

D. ROSENTHAL,* Brussels, Belgium. I 
should like first to express my deep apprecia 
tion to the authors for having given such con 
clusive evidence that no simple diffusion law 
can explain the process of oxidation of copper 
at low temperatures. Similar experiments 
have been performed by Mr. Lucaisi and myself 
at the University of Brussels. We have ex 
amined samples by means of electron diffrac 
tion in a device designed by Professor Finch, 
of the Imperial College of Technology, at 
London. 

Our samples were of electrolytic copper and 
they were given the usual metallographic 
polishing. They were heated in vacuum up to 
900, then they were examined by the dif 
fraction X-ray method. We found that the 
superficial layer was still in an amorphous 
state, as always after polishing. Despite very 
prolonged and very high-temperature anneal 
ing, the amorphous layer could not be removed; 
it could be removed only by etching, but we 
have found that etching renders more difficult 
the subsequent examination of oxide layers. 

When subjected to oxidation at 200 C., 
samples yielded after 3 min. a layer whose 
structure was the usual structure of cuprous 
oxide, but the size of the unit cell was 10 per 
cent greater, as was to be expected. We made 
the assumption that oxygen atoms are in solid 
solution in cuprous oxide and this would 
account for the increase of the unit cell. 

Examination of samples after more prolonged 
heating, i hr. and up to 48 hr. 3 showed progres 
sive changes in the structures. The cuprous 

* Chef de Travaux et Agrege de L Universite^de 
Bruxelles. 



2 JO 



LOW-TEMPERATURE OXIDATION OF SINGLE CRYSTALS OF COPPER 



oxide type of structure had entirely disappeared 
and had been replaced by a very indefinite 
structure. The image was very poor and the 
lines very diffused, but after more prolonged 
heating the so-called cupric (CuO prime 
structure) appeared and then it was replaced 
by the regular CuO structure. A sample treated 
in the same way and oxidized for only 3 min. 
at 3ooC gave at once the structure CuO 
prime. We believe that if we could have per 
formed the oxidation at lower temperatures 
for instance, under 5oC. like those used by 
the authors of the present paper, we would 
have been able to check perhaps the modifica 
tion of the Cu 2 O structures. I mean that per 
haps we could account for the increase in the 
size of the unit cell. 

It must now be said that the observed struc 
ture applies only to the superficial layer, since 
electron diffraction accounts for only 50 to 
100 A. thickness and ignores totally what 
happens in the layer that is below. Taking all 
this into account, we have now a very fair 
explanation of why a simple diffusion law can 
not explain the oxidation of copper at low 
temperatures. 

J. B. AUSTIN,* Kearny, N. J.- This paper 
is very interesting because it represents an 
application of the polarographic method, which 
has been rather too much neglected. It is a 
powerful tool and should be used far more 
than it has been. 

Another point of interest is the effect of the 
roughness of the surface on the rate of oxida 
tion. We have been making some measurements 
of the rate of oxidation of iron and steels at 
high temperature and have found that in this 
case the roughness of the surface has a very 
pronounced effect on the initial rate of oxida 
tion at i iooF. The difference between a milled 
surface, a pickled surface and a polished 
surface is noticeable, the milled surface oxidiz 
ing most rapidly and the polished surface least 
rapidly. This effect of the condition of the 
surface on the rate of oxidation has not been 
given as much attention as it deserves. 

S. L. HoYT,f Columbus, Ohio. In a study 
that I happened to make in this connection I 
was interested in the rate of oxidation of a 

* U. S. Steel Corporation, Research Laboratories. 
t Technical Adviser, Battelle Memorial Institute. 



copper surface. The only thing of importance 
that seems worth mentioning here is that the 
particular experiment that I was running 
seemed to be interfered with by the evolution 
of gas from the copper sample. The method 
used was extremely sensitive to changes in 
pressure from this source. 

In view of the anomaly reported by the 
authors at temperatures between 105 and 
i3oC., it seems possible that the evolution of 
gas from the sample may have played some 
part. If the authors were studying the forma 
tion of scale, I do not believe this effect would 
have any bearing, but with a film of quite small 
thickness, the evolution of various gases, 
nitrogen, carbon monoxide, hydrogen and so 
forth, might in some way interfere with the 
normal diffusion of oxygen through this film of 
copper oxide, and in that way alter the rate 
at which the film forms. 

The evolution of gas from the sample would 
be, of course, a temperature function and in 
that way would give two temperature func 
tions, one of evolution of gas and the other of 
film formation. This suggests the use in these 
experiments of a thoroughly degasified sample 
of copper. 

B. LUSTMAN AND R. F. MEHL (authors 1 
reply). As Dr. Evans states, all methods of 
determining film thickness by optical reflection 
from a film-covered metal surface (including 
that used in the present investigation) measure 
the distance between the film-gas and film- 
metal interface. In the method of surface 
preparation employed here, where smooth, 
single-crystalline, pure copper was used, the 
amount of "interlocking" or of internal oxide 
should be very slight, if not nonexistent. 

While certainly, as Dr. Vernon points out, 
studies on the rate of oxidation of rough 
abraded surfaces are of considerable practical 
interest, it was felt that a full understanding 
of the external oxidation process would require 
studies in the first instance on surfaces as 
ideally smooth as possible. 

The change-over from the logarithmic equa 
tion to the parabolic equation may well vary 
with the manner of surface preparation. It 
would be of interest to determine whether there 
is a transition temperature for the change-over 
from the one equation to the other or whether 
this occurs at a transition film thickness. 



DISCUSSION 



271 



It seems very doubtful indeed that the first- 
formed oxide on copper is amorphous; the fact 
that a well-defined orientation relationship 
exists between oxide and metal even when the 
former is very thick 50 would seem to disprove 
this contention conclusively, for surely an 
amorphous layer of only a few atoms in thick 
ness would be sufficient to prevent an orienta 
tion relationship since the force field of an atom 
is inappreciable beyond a few atoms diameters. 
It would, furthermore, seem unlikely that the 
oxide film would be discontinuous and granular 
on the smooth, strain-free surfaces used in these 
experiments; such a film is more to be expected 
on severely stressed, abraded surfaces. It 
should also be said that brilliant interference 
tints were obtained on copper and iron through 
the range of logarithmic oxidation. The appear 
ance of interference tints would seem to be 
correlated more with surface geometry than 
with manner of oxidation. 

In reply to Dr. Rosenthal s discussion, it 
seems to us very unlikely that an amorphous 
layer (in the usual sense of this term) could 
remain on the copper surface after annealing 
at a temperature of 9ooC. The reasoning 



applied above to the suggestion that the oxide 
may be amorphous may be applied here also. 
It would seem more likely that the surface has 
become exceedingly smooth after this treat 
ment; the halo diffraction pattern character 
istic of amorphous surfaces may result on 
reflection from very smooth surfaces. 51 - 52 
While we agree with Dr. Rosenthal s statement 
that simple diffusion laws will not explain the 
course of oxidation of copper, we doubt, for the 
reasons stated in the paper, that measurable 
changes of phase in the film can account for the 
results obtained. 

It is gratifying to learn that Dr. Austin 
observed an effect of surface geometry in the 
scaling of steel similar to that noted here in 
the rate of oxide-film formation on copper. 

Pretreatment of the copper specimens both 
in a vacuum and in a purified hydrogen atmos 
phere were attempted and the oxidation rates 
in both cases were very similar. It would thus 
appear doubtful that, as Dr. Hoyt suggests, 
evolution of gases from the metal played an 
important role in determining the oxidation 
rates obtained. 



so R. F. Mehl, E. H. McCandless and F. N. Rhines: 
Nature (1934) *34) 1009. 



51 P. Kuchner: Nature (1932) 129, 545. 

52 L. H. Germer: Phys. Rev. (1933) 43, 724. 



Preparation and Some Properties of High-purity Copper 



BY J. S. SMART, JR.,* A. A. SMITH, JR.,* AND A. J. PHILLIPS* 

(New York Meeting, February 1941) 



THE preparation of a sufficient quantity 
of high-purity copper for extensive labora 
tory use in the study of impurity effects 
has been an integral part of a research 
program in progress at the Central Re 
search Laboratory of the American Smelt 
ing and Refining Co. for the past several 
years. Initially the degree of purity desired 
was that necessary to meet two essential 
requirements for experimental accuracy: 
(i) no impurity was to be present in the 
copper in an amount capable of producing 
a detectable effect on the properties to be 
studied, (2) even the minor changes in 
properties caused by the addition of ex 
tremely small amounts of individual ele 
ments were to be measured without 
significant interference from contaminants. 
In the course of the investigation, evidence 
has accumulated to justify the conclusion 
that this copper is of exceptional purity; 
in fact., no impurity can be detected by any 
chemical method or spectrographic tech 
nique known to the authors. The present 
paper describes the method of production 
of the copper and presents some of its 
properties. 

PREPARATION or HIGH-PURITY COPPER 
The elimination of impurities from com 
mercial electrolytic copper is accomplished 
by a three-step purification process. The 
first operation consists of an electrolysis 
through a purified CuS0 4 -H 2 S0 4 electro 
lyte using bagged anodes, low current 
density (5 amp. per sq. ft.) and low tern- 
Manuscript received at the office of the Institute 
N ov. 29, 1040- Issued as T.P. 1289 in METALS TECH 
NOLOGY, February 1941. 

* Central Research Laboratory, American Smelting 
& Refining Co., Barber, New Jersey. 



perature (25C. max.). The electrolyte is 

prepared from commercial Cu 2 S0 4 and is 

purified by heating the solution and adding 

i gram Fe as Fe 2 (S0 4 ) 3 to each 100 grams 

dissolved copper; sufficient hydrated oxide 

of copper, prepared from Cu 2 S0 4 and 

NaOH, is supplied to precipitate Fe(OH) 3 , 

which carries down arsenic, antimony, 

tin, selenium, tellurium and other minor 

constituents. Silver and remaining traces 

of lead are removed by addition of Na 2 S to 

the acidified filtrate, which is then refiltered 

after standing hot overnight. The cold 

solution is made up to 30 grams per liter 

Cu and 150 grams per liter H 2 S0 4 and 

introduced without addition agents into 

the electrolytic system, which consists of 

a double row of hard-rubber battery boxes 

with the inside webbing and partitions 

removed, arranged in seven steps of 

ascending height. Electrolyte is uniformly 

supplied to the two highest cells by glass 

siphons operating from a constant-level 

reservoir, and flows by gravity through the 

bank of seven cells, entering each at the 

bottom and leaving at the top. Glass air 

Lifts continuously supply the returning 

liquid to the constant-level reservoir for 

recirculation. 

Each cell contains three 4% by 4% by 
i-in anodes cast from high-quality electro 
lytic wirebar copper, various lots of which 
have been used with equal success. The 
anodes are pinned to sheet copper hangars 
shaped to prevent crawl of the electrolyte 
to the multiple-wired bus bars which 
would result in fouling of the contacts 
and hang inside wool flannel bags stretched 
over open wood frames. The deposit 



272 



J. S. SMART, JR., A. A. SMITH, JR. AND A. J. PHILLIPS 



2 73 



is obtained on starting sheets previ 
ously stripped from uncoated stainless- 
steel blanks; the finished cathode weighs 
about 5 Ib. All impurities are satisfactorily 
removed except sulphur, which is present 
in the form of entrapped sulphates. 

Despite certain disadvantages, oxidation 
by blowing the molten metal has proved 
the most satisfactory method for removal 
of sulphur. Thirty-pound melts are surface- 
blown to an oxygen content of about 1.5 
per cent in a clay-graphite crucible heated 
by induction. Contamination of the melt 
by clay, iron and graphite is unavoidable, 
but the sulphur content is uniformly 
reduced below the limit of qualitative 
detection by conversion to E^S and absorp 
tion in CdCl2 solution. This degree of 
sulphur elimination is essential, since it is 
not completely removed by the subsequent 
processes. The melt is cast in the form of 
anodes, which are then re-electrolyzed 
through a high-purity Cu(N0 3 )a solution 
to remove the contaminants picked up in 
the blowing operation. This electrolysis is 
carried out in glass cells at room tempera 
ture, using a current density of 15 amp. per 
sq. ft. The anodes are of the previously 
described dimensions and the starting 
sheets are likewise deposited in the appara 
tus on stainless-steel blanks. The electro 
lyte is made from pure copper and HNOs 
and is periodically purified by addition of a 
large excess of NH 4 OH to precipitate the 
accumulated iron. After nitration, sufficient 
HNOs is added to precipitate most of the 
copper at pH 4, and the precipitate is 
washed by decantation to eliminate the 
bulk of the NEUNOs. A minimum of 
HN0 3 is used to redissolve the precipitate, 
which is then diluted to 25 grams per liter 
Cu and used without further treatment 
at a pH of 2. The final deposit is bright, 
dense and very pure except for nitrates, 
which are eliminated by subsequent 
melting. 

Inasmuch as the experimental work is 
generally performed on drawn wires, the 



final manipulations involve the production 
of an oxygen-free pure copper rod suitable 
for drawing. Fortunately, surface con 
tamination from steel tools and other 



N 2 exhaust 



inlet 




Drive rolls 



FIG. i. SECTION OF LABORATORY CONTINUOUS 

CASTING APPARATUS. 

similar sources can be removed by pickling 
in HN0 3 , followed by hot 1:1 HC1, a 
precaution that is observed prior to all 
melting operations. The choice of a crucible 
that introduces no impurities into the 
melt and is adaptable to the production of 
oxygen-free castings is of great importance. 
Most crucibles contain either iron or 
sulphur in objectionable amounts, or are 
too porous. A special high-purity grade of 
Acheson graphite is the only material 
known to possess the necessary properties. 
It is available in the form of 4-in. diam. 
electrodes, from which the desired crucibles 
are machined for each of the various 
melting operations. Oxygen-bearing melts 
may be readily cast from an open crucible 
by pouring through air into a mold of the 
same material. Pure oxygen-free copper 
and its alloys are cast in the form of 



PREPARATION AND SOME PROPERTIES OF HIGH-PURITY COPPER 



smooth, dense ?g-in. rod by a continuous 
casting process in which the cooling is 
effected laterally. 

Fig. i represents a diagrammatic cross 
section of the necessary apparatus. A 
previously prepared lo-lb. oxygen-bearing 
pure copper slug is accommodated by 
crucible A, which rests on insulating ring 
F. Tight-fitting cover B is provided with 
pure graphite tubes for the inlet and 
exhaust of purified nitrogen sweeping gas, 
the whole being packed in tamped lamp 
black C retained by silica sleeve D. The 
graphite casting die G is threaded through 
the bottom of the crucible and is encased 
in a water jacket /, which extracts prac 
tically all of the heat laterally through the 
copper and die wall. After the casting die 
is plugged with a pure copper " starting 
rod," which is held in position by the 
tension on the driving rolls, heat is sup 
plied by means of a high-frequency coil E. 
The bottom of the crucible is protected 
from oxidation by a hydrogen pocket, and 
the melt is allowed to stand for about 
30 min. while deoxidation and thorough 
mixing are accomplished. The nitrogen 
sweeping gas removes excess hydrogen to 
permit the production of high-density 
copper rod. The power-driven rolls auto 
matically cast the melt at a uniform rate by 
continuous withdrawal of the solidified rod 
from the die. 

This process is ideal for the elimination 
of oxygen, because both solidification of 
the melt and further cooling of the rod to a 
low temperature occur within an enclosed 
system where the copper is in contact at 
all times with only carbon, carbon mon 
oxide, hydrogen and nitrogen. However, 
since strictly oxygen-free copper was 
considered essential, experimental verifica 
tion of the efficiency of deoxidation was 
obtained by several methods. The evidence 
is necessarily of an indirect nature because 
the accepted method for analysis of 
oxygen is definitely known to be inaccurate 
for extremely small amounts, owing to a 



variable adsorption of oxygen on the sur 
face of the sample. 1 The addition of 
0.00007 per cent Fe to pure copper resulted 
in a small but reproducible loss of con 
ductivity, indicating that insufficient oxy 
gen was present to remove this amount of 
iron from solid solution by formation of 
an insoluble oxide. According to Rhines, 2 
the composition of the oxide is Fe 3 4 . If 
it is assumed that the quantity of dissolved 
oxygen in equilibrium with solute iron is 
negligible, the presence of only 0.000027 
per cent Os would have completely pre 
cipitated the iron and eliminated the 
conductivity drop. Another and more 
striking example of the extent of deoxida 
tion was supplied by adding Si0 2 to a 
charge of pure copper, which was then 
melted and continuous cast in the usual 
way. Reduction of Si0 2 was effected, the 
silicon entering solid solution to form an 
alloy. The same alloy removed traces of 
oxygen from commercial hydrogen at 
85oC. by formation and precipitation of 
Si0 2 in the copper. Hydrogen, in turn, 
readily removes oxygen from pure copper 
at the same temperature. 

The employment of the usual chemical 
methods for the detection of minute 
amounts of other impurities, especially 
those normally present in commercial cop 
per, has proved generally unsatisfactory 
because of a lack of sensitivity. The spec- 
trograph offers more attractive possibilities 
for work of this type, particularly when a 
high-arcing current is used across suitable 
lengths of rod acting as self-electrodes. The 
limits of visibility have been explored on 
alloys prepared by synthesis from pure 
copper and various master alloys. Despite 
the excellent sensitivity of the method, no 
impurity is detectable in the cast oxygen- 
free pure copper rod. Although a precise 
statement with regard to specific purity is 
not possible at the present time, it is certain 
that the minimum copper content is 99.999 
per cent. 

1 References are at the end of the paper. 



J. S. SMART, JR., A. A. SMITH, JR. AND A. J. PHILLIPS 



275 



Column II of Table i lists the composi 
tion of binary alloys of pure copper and the 
common impurities. Each of the impurity 
contents noted is detectable spectrograph- 
ically and represents a definite excess over 
the quantity present in pure copper. Most 
of the compositions given in that column do 
not correspond to the low limits of spectro- 
graphic detection that have been conserva 
tively estimated from the line gradients and 
listed in column III. In all probability the 
amount of each impurity present in the 
pure copper is indeterminately lower than 
that represented in the latter column. 

TABLE i. Impurity Limits in Continuous 

Cast High-purity Copper Rod 

PER CENT BY WEIGHT 



I 


II 


III 


IV 


Ele 
ment 


Definite 


Probable 


Method 


Fe 


<o. 00007 


<o. 00005 


Spectrographic 


Sb 


<O.OOOI 


<O.OOOI 


Spectrographic 


Pb 


<O.OOOI 


<o. 00005 


Spectrographic 


Sn 


<O.OOOI 


<o. 00005 


Spectrographic 


Ni 

Bi 


<O.OOOI 
<O.OOOOI 


<O.OOOI 
<O.OOOOI 


Spectrographic 
Spectrographic 


AS 

As 


<o. 00003 

<0.0002 


<O.OOOOOI 
<O.OOOI 


Spectrographic 
Spectrographic 


Cr 


<o. 00005 


<O.OOOOI 


Spectrographic 


si 


<O.OOOOI 


<O.OOOOOI 


Spectrographic 


Te 


<O.O002 


<O.OOOI 


Spectrographic 


Se 


<O.OOOI 




Chemical, 500 grams 


S 


<0 0001 




Chemical 



Checked by E. N. Skinner, Jr. Unpublished 
Thesis to Faculty of Yale University, May 1940. 

From the description of the casting 
method, it is apparent that several other 
elements are in close contact with the melt 
and should receive consideration as con 
taminants; viz., C, CO, N 2 and H 2 . The 
solubility of carbon monoxide in solid 
copper has been demonstrated to be of a 
negligible order, by Sieverts and Krumb- 
haar, 3 Rontgen and Muller, 4 Allen, 5 and 
Ransley. 6 This, however, does not preclude 
lack of solubility on the part of carbon, 
although the possibilities are decidedly 
limited theoretically. If for the moment a 
small solubility is assumed, which is in all 
probability dependent on temperature, 
changes in certain properties would be 
expected from various heat-treatments 



designed to produce the solubility ex 
tremes. The results of experimental ex 
ploration of these possibilities have been 
entirely negative. Furthermore, the addi 
tion of oxygen should prove beneficial (as 
previously described for iron) if carbon is 
present in sufficient amounts to affect the 
properties, and this is not true. 

Hydrogen is undoubtedly present in the 
cast rod, and the quantity in solution is at 
least partly dependent on the efficiency of 
the previously described nitrogen sweep. 
Although vacuum removal has not been 
attempted, samples of the rods have been 
surface-oxidized in air at 85oC. and heated 
at this temperature in an atmosphere of 
pure dry nitrogen for periods as long as 2 1 
days. This treatment should result in a 
substantial diminution of the hydrogen 
content for comparison with other samples 
that were saturated with hydrogen at the 
same temperature. Koster 7 extracted hy 
drogen from several types of copper at 
elevated temperatures, cold-worked the 
samples, and found a substantial decrease 
in the temperature at which softening 
occurred as compared to the original 
untreated material. However, none of the 
pure copper samples prepared as above 
exhibited any trace of this effect; the values 
obtained for the original cast rod, the 
oxidized and nitrogen-swept rod, and the 
hydrogen-saturated rod all agreed within 
the limits of accuracy of the method. It is 
the authors opinion that the changes in 
properties noted by Koster were due to 
variations in the solubility of impurities 
other than hydrogen, resulting from the 
thermal treatment involved in the vacuum 
elimination. No evidence has been obtained 
to justify the proposition that hydrogen 
has an influence on the physical properties 
of pure copper in the solid state. 

The many published accounts regarding 
the insolubility of nitrogen in both molten 
and solid copper will not be reviewed here. 
All of the present experimental observa 
tions confirm this generally accepted 



276 



PREPARATION AND SOME PROPERTIES OP HIGH-PURITY COPPER 



premise. If a finite solid solubility of nitro 
gen in copper actually does exist its effect 
on the properties that have been studied is 
indeterminately small. 

SOME PROPERTIES OF HIGH-PURITY COPPER 

As cast, the M-in. rod is very soft and 
evidently can withstand an unlimited 
amount of cold-working, but the grain size 
is undesirably coarse for many purposes, 
and a number of intermediate anneals are 
usually employed in the cold-drawing 
schedule to obtain refinement. Variations 
in the annealing temperature between 300 
and 8ooC. have a negligible influence on 
the properties investigated, exclusive of 
grain size and its related effects, as would 
be expected from a metal of high purity. 

Conductivity 

The conductivity of numerous lots of 
pure copper as determined by a standard 
Hoopes bridge over a period of several 
years averages 102.3 per cent of the 
I.A.C.S., corresponding to a resistivity of 
0.14983 ohms per meter gram at 2oC. As 
would be expected from the experimental 
limits of error, the determinations vary 
between 102.25 and 102.35 per cent. 
Theoretically, oxygen present in solid solu 
tion should lower the conductivity of 
strictly pure copper, but its addition to 
impure copper invariably results in a con 
ductivity increase due to removal of certain 
impurities from solid solution. The effect 
of soluble oxygen has been investigated by 
permitting various samples of oxygen-free 
rod to absorb oxygen by diffusion to the 
point of saturation at 85oC., and drawing 
test wires for measurement. Table 2 lists a 
number of determinations made on a stand 
ard Kelvin bridge, which indicate that 
oxygen does actually lower the conductivity 
to a very small extent. The oxygen-free 
samples were annealed in hydrogen at 
5ooC. and oxygen-bearing wires in nitro 
gen at 500 and 85oC. The samples in the 
form of straight lengths were immersed in 



an oil bath of uniform temperature and the 
measurements were corrected to 2oC. by 
the usual method. 

TABLE 2. Conductivity of Samples An 
nealed for One Hour at Indicated Tem 
peratures and Cooled by Rapid 
Quenching 
PER CENT 



Sam 
ple 
No. 


Description 


Conductivity, 
Per Cent 


An 
nealed 
at sooC. 


An 
nealed 
at8soC. 


A 
B 

C 

D 

E 


Oxygen-free 


102.32 
102.23 

102.21 

102.39 
102.19 


102.10 

102,23 


Oxygen-bearing (satu 
rated at 8soC.) 


Oxygen-bearing (check 
run) 


Oxygen-free. . . . 


Oxygen-bearing 





a Samples D and E taken from the same wire as 
samples A and B were kindly measured by the investi 
gators of another laboratory. 

Because of the small magnitude of the 
loss of conductivity as measured from the 
wires annealed at 500 C., where the solu 
bility of oxygen is considerably lower than 
at 85oC., the anneals at the higher tem 
perature were included with the expectation 
of obtaining a larger difference. The data 
are inconclusive on this point, but the 
results cannot be definitely classified as 
anomalous because of lack of information 
on their approach to equilibrium condi 
tions, and the limits of experimental ac 
curacy. There is no disagreement, however, 
with respect to the existence of a loss of 
conductivity due to the presence of oxygen 
in solid solution. 

Cold-working 

The effects of cold-work on the mechan 
ical properties of copper, and the factors 
influencing its annealing characteristics 
have been studied by numerous investiga 
tors. For the most part the data now avail 
able have been obtained from coppers of 
varying commercial composition, and since 
impurity effects are often of considerable 
magnitude, the variations in the properties 



J. S. SMART, JR., A. A. SMITH, JR. AND A. J. PHILLIPS 



277 



of high-purity copper when cold-worked 
and annealed under certain conditions are 
of general interest. Some effects of cold- 
working \ull be considered first, following 



tested on a Scott wire tester of 5o-lb. 
capacity at an indeterminate higher rate 
of strain, the minor disadvantages of which 
were more than overcome by the increased 



ou 
_c 

S_ 
CD 
CL 










^^ 


- 




DU 

40^ 

30| 

S. 
?0| 

10 J 

LU 








^ 










Conductivity, percent l.A C. 
S 8 2 g j 
Tensile strength, 1000 1 

>O OJ - cn d 
C 


/ 


X^ 


> 
























Elongt 
50 


Us 

75 


ISjjl 










Redud 
87.5 


on, per 
95 


cent 
98.1 


994 





12 16 20 24 28 
Reduction . B and S. No 
FIG. 2. EFFECT OF COLD-DRAWING ON SOME PROPERTIES OF PURE COPPER. 



which they will be compared with the 
changes accompanying annealing. 

The cast structure of the %-in. con 
tinuous cast rod was broken down by cold- 
drawing 30 per cent, 32 per cent and 37 per 
cent, employing an intermediate anneal of 
30 min. at 6ooC. in hydrogen after each 
of these reductions, to provide 0.204 in. 
(No. 4 B. and S.) annealed wire (0.07 mm. 
grain size) for the determination of the 
effect of deformation by cold-drawing on 
the tensile strength, elongation and con 
traction of area. The wire was then cold- 
drawn 22 B. and S. numbers on a string-up 
machine at a rate of 192 ft. per min., being 
coiled on a block of 17-111. diameter and 
cooled between dies. Test samples were 
taken after most of the die reductions, as 
permitted by the quantity of the wire. 

Tensile strength and elongation were 
determined from wires of lo-in. gauge 
length; the diameters were measured by 
means of micrometers. For samples drawn 
up to 13 numbers, two recently calibrated 
hydraulic machines were used, which 
could be read to within i.o per cent of the 
applied load, using a rate of strain of 0.0125 
in. per in. per min. The finer sizes were 



sensitivity of load measurement. Elonga 
tion measurements were taken from divider 
settings, using small scratches on the 
specimens for reference markings. Because 
of the small size of the samples, the de 
terminations of the contraction of area are 
subject to sufficient error to mask the small 
changes that occur during the cold- working 
of this very pure material, even though 
the fractures were magnified 50 X and 
projected on a ground glass for evaluation. 
In general, a contraction of area of 94 per 
cent, representative of the annealed wire, 
dropped to about 92 per cent for samples 
drawn 14 numbers. 

The conductivities of samples cold- 
drawn o to 9 numbers were also determined, 
each of the measurements being made on 
0.08 i-in. diam. wire using a standard 
Hoopes bridge. These samples were drawn 
from a hot-rolled rod fabricated from an 
oxygen-free slug in such a way as to 
minimize exposure to air at high tempera 
ture, and consequent oxygen absorption. 
The rod was then drawn through the 
various sizes between 0.229 and 0.08 1 in. 
and a sample from each die reduction 
was annealed at sooC. for i hr. This 



278 



PREPARATION AND SOME PROPERTIES OP HIGH-PURITY COPPER 



procedure furnished a series of annealed 
wires of progressively increasing diameters, 
but of similar grain size (0.035 to 0.04 mm.), 
which when cold-drawn to 0.08 1 in. pro 
duced samples drawn o to 9 numbers 
hard. 

The comparative effects of cold deforma 
tion by drawing on tensile strength, 
elongation and conductivity are repre 
sented in Fig. 2. By plotting the data 
against B. and S. numbers, each of which 
represents a reduction of slightly over 
20 per cent of the cross-sectional area of 
the wire entering the die, the progressive 
effects of equal increments of reduction 
may be determined directly from the 
curves. Thus the first four reductions 
have large but steadily decreasing unit 
effects on the tensile strength, after which 
the relationship becomes linear for the 
next 14 numbers. At higher reductions a 
further decrease in the slope of the curve 
might be predicted on the basis of an 
appreciable increase in the tendency for 
self-annealing, but the small dimensions 
of the wires introduce errors of such magni 
tude as to preclude accurate determination 
of this possibility. 

Alkins 8 has described an arrest in the 
increase in tensile strength of hard-drawn 
copper, but has plotted his results against 
cross-sectional area, and the curve does 
not represent the effects of equal incre 
ments of cold reduction. Although the 
reduction at each stage of the drawing 
procedure was not the same for all of 
Alkins test wires, the six samples that 
determined the arrest had been reduced 
13.6 per cent per draft, with one exception 
of 12.6 per cent. When replotted against 
reduction numbers, these data did not 
indicate the presence of the arrest, but 
determined a straight line, the deviations 
being within the limits of error and of 
about the same order as the data of Fig. 2. 
A similar ambiguous effect is obtained by 
plotting tensile-strength data against total 
percentage reduction of area. 9 " 11 



Of particular interest^ is the manner in 
which the various properties change with 
the degree of cold-working. Fig. 2 clearly 
indicates that elongation is the property 
first affected; the change occurs so early 
that the major part of the total decrease is 
obtained during the first die reduction. 
The next highest initial rate of change is 
to be found in the conductivity curve. 
Impurities, in small amounts, do not appear 
to have a significant effect on the total 
loss of conductivity caused by cold-work 
ing, since the average loss of a number of 
wirebar coppers drawn 9 numbers was 
2.25 per cent as compared to the present 
drop of 2.3 per cent. Similarly, the tensile 
strength of annealed pure copper differs 
very little from that of the commercial 
product; both appear to vary between 
31,000 and 32,000 Ib. per sq. in. The rate 
of increase of the tensile strength, as 
determined by the first increments of 
cold-work, is decidedly smaller than the 
similar changes of opposite sign occurring 
in the elongation and conductivity. 

One of the most interesting effects of 
cold deformation on the properties of 
copper is the well-known decrease in 
density obtained from wrought and an 
nealed coppers that bave been further 
cold-worked by any of several methods. 
Typical measurements of the density 
changes as functions of the degree and 
method of cold- working have been reported 
by Alkins 10 and Maier. 12 In addition, there 
appears to be a discrepancy between the 
density of copper as determined directly 
and that calculated from X-ray measure 
ments. Maier obtained a value of 8.95285 
grams per c.c. at 2oC. from a single 
crystal, which is considerably higher than 
the calculated result of 8.9381 grams per 
c.c. tabulated by Stockdale 13 from the 
X-ray determinations of Van Bergen 14 
and Roberts. 15 Although the scope of 
the present paper does not permit a 
thorough study of these phenomena, it 
was thought advisable to include sufficient 



J. S. SMART, JR., A. A. SMITH, JR. AND A. J. PHILLIPS 



279 



data to indicate the behavior of pure 
copper when subjected to various treat 
ments that might be expected to alter its 
density. 



continuous cast rod were measured in the 
as-cast condition, and the remainder cold- 
drawn by the method previously described 
for further tests. 



TABLE 3. Averaged Densities in Grams per Cubic Centimeter at 20 C. 



Nominal 
Reduc 
tion, Per 
Cent 


Oxygen-free Pure Copper 


Pure Copper 
+ 0.037 Per 
Cent O 2 


Commercial Copper 


i 


2 


3 


4 


5 


6 


7 


Single 
Crystal 


Coarse Poly- 
crystalline 


Fine Poly- 
crystalline 


Fine Poly- 
crystalline 


Wirebar 
0.041 Per 
Cent Oa 


Alkms 
Data 


Maier s 
Data 


o 
So 
90 


8.9577 


8 9592 
8.9560 
8.9550 


8.9581 
8.9563 
8.9548 


8.9379 
8.9304 
8.9214 


8.9314 
8.9200 
8.9226 


8.9165 
8.8918 
8.8866 


8.92426 
8.90877 
8.91052 



DIFFERENCES IN DENSITY CAUSED BY COLD-DRAWING 



so 
90 



0.0032 
0.0042 



0.0018 
0.0033 



0.0075 
0.0165 



0.0114 
0.0088 



0.0247 
0.0299 



0.01549 
0.01374 



Actual reductions for rod No. 2 were 50.5 and 90.2 per cent; for Nos. 3 to 5, 49-8 and 90.0 per cent; for No. 
6, 47.29 and 89.76 per cent; for No. 7, 53-37 and 90.53 per cent. 



The effects of grain size and several de 
grees of deformation were determined from 
pure copper in the oxygen-free condition, 
and compared with the results obtained 
from pure copper and a typical sample of 
commercial electrolytic copper of wirebar 
grade, both of which contained excess 
oxygen in the form of cuprous oxide. A 
single crystal of pure copper was measured 
in the cast condition but unfortunately 
was not large enough for deformation by 
drawing. Polycrystalline material con 
tinuously cast from the same lot provided 
samples having a large grain size, while 
hot-rolled pure copper rod furnished the 
fine-grained condition. The latter sample 
was hot-rolled from an oxygen-free slug 
by the previously noted method of short 
exposure to air at high temperature. The 
oxygen-bearing pure copper rod was hot- 
rolled from a cast slug, and the commercial 
sample from a wirebar slice. All hot-rolled 
rods were drawn from 0.3125 to o.ssy-in. 
diameter and annealed i hr. at 6ooC. to 
furnish annealed samples for measurement 
and further cold-working. Portions of the 



Density measurements were made on 
carefully cleaned portions cut from each 
end of the annealed or cast rods and on 
samples from the remaining sections after 
cold-drawing 50 and 90 per cent. The 
method employed is similar in principle 
to that described by Maier. Samples of 
15 to 50 grams were weighed on a large 
analytical balance sensitive to 0.0001 
grams, and then in C.P. CCU, in which 
they were suspended by means of a fine 
platinum wire. A temperature-density 
curve for the CQU was constructed by use 
of a picnometer, which in turn was cali 
brated against mercury. The CCU was 
contained in a glass flask of 5oo-ml. capac 
ity, equipped with a cork that admitted 
the wire through a small glass tube. The 
volume of the flask was large enough so 
that by stirring immediately prior to 
weighing, and suspending the sample in 
the center of the container, temperature 
gradients were eliminated within the limits 
of sensitivity of the thermometer, which 
could easily be read to o.o5C. The same 
thermometer was used for all measure- 



2 80 



PREPARATION AND SOME PROPERTIES OF HIGH-PURITY COPPER 



ments, including the determination of the 
density of the CCU, and occupied a 
position immediately adjacent to the 
specimen in the flask. As a precaution, the 
apparatus was generally allowed to remain 
undisturbed for a sufficient period of time 
after the initial weighing to ensure a 
change in temperature, then a second 
reading was made; the checks were always 
within, and generally much better than 
5 parts per 100,000. After appropriate 
deductions for the suspension wire, the 
measurements were corrected to 2oC. on 
the basis of a volume coefficient of expan 
sion for copper of 3 X 16.8 X io~ 6 per 
degree centigrade. The accuracy of the 
method is conservatively estimated as i 
part in 10,000. The results are listed in 
Table 3. 

It is evident that the densities obtained 
from the oxygen-free pure coppers are 
somewhat higher than Maier s results for a 
single crystal, and that the order of the 
discrepancy between these values and that 
calculated from lattice-parameter measure 
ments is far too large to be dismissed as an 
error in the experimental measurements. 
Since the density of the single crystal is 
slightly lower than the polycrystalline cast 
rod made from the same lot, and approxi 
mately equal to that of the hot-rolled rod, 
it is most probable that grain size has such a 
small effect on density that its determina 
tion is beyond the accuracy of the present 
method, including the variation in density 
encountered in the preparation of the 
samples as part of the error. 

The decreases in the densities of the two 
oxygen-free pure coppers after reduction 
of 50 and 90 per cent are of a very small 
order by comparison with those encoun 
tered in commercial coppers, as found in 
the present investigation, and by Alkins 
and Maier. The aspects of the entire 
phenomenon have been so altered, in fact, 
that it is even doubtful whether pure copper 
actually does suffer a decrease in density 
during cold-working. It is freely admitted 



that the present data do no more than 
suggest this possibility. Its proof is quite 
another matter, involving some advanced 
investigation of the effects of the increased 
surface produced by deformation on the 
accuracy of the determinations, and an 
extremely careful research. 

The addition of 0.037 per cent Qz to pure 
copper resulted in a loss in density in 
the annealed condition, which, curiously 
enough, is greater than the effect calculated 
from the density of cuprous, oxide. Further 
more, a subsequent loss due to cold-working 
was obtained, and its magnitude leaves 
little doubt that cuprous oxide is an impor 
tant contributor to this phenomenon, 
possibly by giving rise to the formation of 
minute internal voids. There is considerable 
variation in the behavior of the oxygen- 
bearing coppers with reference to the 
relationship between density loss and 
percentage of reduction, but impurity 
effects must receive consideration, and in 
addition, Alkins and Maier have indicated 
the existence of a minimum in the curve. 

Annealing Characteristics 

The large influence of impurities on the 
softening temperature of copper is well 
known, and there is little question that the 
thermal history prior to cold deformation 
is also an important consideration if im 
purities are present. High-purity copper 
therefore greatly facilitates the acquisition 
of fundamental data concerning the effects 
of three important variables deformation, 
annealing time and annealing temperature. 
According to Tammann, 16 the restoration 
of certain properties during annealing is a 
simultaneous process for copper, but since 
it has been shown that the changes occur in 
a definite sequence during cold drawing, it 
is also of interest to compare the results of 
annealing from this standpoint. 

The necessary data were secured from a 
lo-lb. coil of %-in. oxygen-free continuous 
cast rod. In order to obtain a series of 
samples reduced 50 per cent, 75 per cent 



J. S. SMART, JR., A. A. SMITH, JR. AND A. J. PHILLIPS 



28l 



and 87.5 per cent by cold-drawing, but of 
o.o8i-in. dia. finished gauge, the break 
down drawing and annealing cycles neces- 



per cent, annealed for i hr. at i4oC. and 
tested in tension. Their tensile strengths 
agreed within 1000 Ib. per sq. in. which, 



60 



y-40 



30 



25101 
?2 



520 



:% 



y 



W 6(T 80 100 120 140 160 180 200 
Anneoiling temperature, deg.C. 

FlG 3 _ ANNEALING CHARACTERISTICS OF PURE 

COPPER WIRE REDUCED 50 PER CENT. 

sarily differed for each series, and are given 
in abbreviated form: 

50 Per Cent Series: reductions of 53.5 
per cent, 37 per cent, 37 per cent and 37 per 
cent, each followed by an anneal for 30 
min. at 6ooC. Finish-drawn 50 per cent 
to 0.081 inch. 

75 Per Cent Series: reductions of 30 per 
cent, 32 per cent, 37 per cent and 37 per 
cent, each followed by an anneal for 30 
min. at 6ooC. Finish-drawn 75 per cent 
to 0.081 inch. 

87.5 Per Cent Series: reductions of 40.5 
per cent and 37.5 per cent, each followed by 
an anneal for 30 min. at 600 C. Finish- 
drawn 87.5 per cent to 0.081 inch. 

After the final intermediate anneal, small 
samples of each of the various rods were 
tested for uniformity with respect to their 
response to a standard annealing treatment 
to ensure freedom from such variations as 
might have been produced by the differ 
ences in the break-down procedures. For 
this purpose the samples were reduced 75 



DU 






1 










- 

-*- rj> rr\ 




"X 




*V 


V 






cp50 

L? *J 

-t Q. 

w5 . /n 




\ 


^r 


\ 


t S 


u^ 

^ 




,u^ 4U 
~5o 

CO 
^ 7H 






%, 




*%i 


V 





















1/5 109 






^ 






x 




^O 102 

>-< 
^^ 

irj-t: IAI 






f 




/ 


/ 




-p 01 
cS o3 




J 




..y 


<.s 


> 




^inn 




u u 




























. 
















2 4 

OVt OA 






r 




[/ 






C ZU 

_o <-> 

LU QJ 

- n 






I 




/, 


/ 






8 


1C 


)0 12 


1 


10 K 


IE 


20 



Annealing temperature, cleg. C. 
FIG. 4. ANNEALING CHARACTERISTICS OF PURE 
COPPER WIRE REDUCED 75 PER CENT. 

from Fig. 4, will be seen to be well within 
the limit of reproducibility for check wires 
subjected to this treatment. Following this 
justification of the method of preparation, 
the processed wires were drawn to test size 
and cut to lengths of i meter each. Samples 
were then annealed in two automatically 
controlled oil baths, one for temperatures 
below i30C., in which the maximum 
variation was o.25C., the other for 
higher temperatures within limits of 
o.sC. For each of the three reductions, 
three annealing periods were used 15 
min., i hr. and 24 hr. Conductivities were 
determined on a Hoopes bridge, following 
which the wires were cut in half to furnish 
check samples for the tensile test. The 
annealing curves are plotted in Figs. 3, 4 
and 5. 

It is evident that during annealing the 
changes in tensile strength, conductivity 
and elongation do not proceed simultane 
ously, but in a sequence which is the reverse 
of that occurring during cold-working. A 



282 



PREPARATION AND SOME PROPERTIES OF HIGH-PURITY COPPER 



comparison of the curves for the i-hr. 
anneals of Fig. 4, along the i4oC. vertical, 
Illustrates a 50 per cent loss of work- 
hardening as measured by tensile strength, 



60 







II 20 



1 



7 



60 80 100 120 140 160 180 POO 
Annealing temperature, deg.C- 

FIG. 5. ANNEALING CHARACTERISTICS OF PURE 
COPPER WIRE REDUCED 87.5 PER CENT. 



a much smaller restoration of conductivity, 
and the appearance of but a slight increase 
in elongation. Additional comparisons 
prove the generality of this effect. 

By inspection of the tensile-strength 
curves, the effects of reduction, time at 
annealing temperature, and the general 
temperature range over which recrystalli- 
zation occurs are self-evident. However, it 
is often of considerable convenience to 
express the softening process in terms of a 
definite softening temperature, for com 
parative purposes. This generally involves 
the determination of a portion of the 
annealing curve by adoption of a standard 
reduction and annealing time, and the 
selection of some point on the curve, such 
as the beginning of the process or the end 
of the process, from which the desired 
temperature may be read. A procedure 
that depends on the determination of the 
half-hard point has several features to 



recommend its selection.* Variations in the 
tensile strength between various coppers 
given equal cold reductions, and differences 
in the temperature range over which soften 
ing occurs, are accounted for, and good 
accuracy is ensured by the large slope of the 
curve as it passes through this range. 

The half-hard stages of the curves of 
Figs. 3 to 5, as determined from the plotted 
annealing data, are designated by the small 
horizontal lines. For example, detectable 
softening of pure copper cold-drawn 87.5 per 
cent and annealed for 24 hr. first appears 
at about 7SC., is complete at about i ioC., 
and the softening temperature is 89C. It 
is of some interest to note the differences 
in softening temperature caused by varia 
tions in the annealing time (Table 4). 

Allowing for the limited scope of the 
data, which does not permit generaliza 
tion, it appears that the effect of annealing 
time is reasonably free from dependence 
on reduction for the particular reductions 
investigated. In addition, the data fit the 
general rule that a decrease in softening 
temperature of ioC. is obtained by doub 
ling the annealing period. 

TABLE 4. Decrease in Softening 

Temperature Due to Indicated 

Variations in Annealing Time 

DEGREES CENTIGRADE 



Reduction, 
Per Cent 


15 Min to i Hr. 


15 Min to 24 Hr 


50 

75 
87-5 


21 
20 
21 


66 
62 
60 



SUMMARY 

i. High-purity copper of a minimum 
purity of 99.999 per cent Cu has been 

* The method extensively used by the authors over 
a period of several years for many different types of 
copper has proved thoroughly reliable. The sample to 
be investigated is reduced 75 per cent by drawing and 
its tensile strength is determined in this, the "hard" 
condition. The annealed tensile strength is obtained 
from a i-hr. anneal at sooC.; these extremes deter 
mine the half-hard point. A sufficient number of i-hr. 
anneals at suitable temperatures provide the data 
necessary to plot the portion of the curve passing 
through the _ half -hard range, and the softening 
temperature is obtained graphically. In most cases 
the reproducibility of the method is 2C. 



DISCUSSION 



283 



prepared in sufficient quantity for exten 
sive use as a research material. 

2. The resistivity of oxygen-free high- 
purity copper is 0.14983 ohm per meter 
gram at 2oC., corresponding to a con 
ductivity of 102.3 per cent of the I.A.C.S. 

3. The presence of oxygen in solid solu 
tion in pure copper increases its resistivity 
by a small but measurable amount. 

4. Data have been presented to indicate 
the changes in tensile strength, elongation 
and conductivity that occur when pure 
copper is progressively cold-worked by 
drawing. 

5. The density of pure copper has been 
determined as 8.9592 grams per c.c. at 

20C. 

6. The density changes that accompany 
the cold-working of oxygen-free pure cop 
per are of a very small if not a negligible 
magnitude. When oxygen is present in 
the form of cuprous oxide, the decrease of 
density during cold-working is appreciable. 

7. The changes in tensile strength, 
elongation and conductivity, of cold- 
worked pure copper annealed at various 
temperatures have been determined for 
three different degrees of reduction and 
annealing time. 

ACKNOWLEDGMENT 

The authors take great pleasure in ex 
pressing their indebtedness to Dr. R.D. 
McLellan, for assistance in the spectro- 
graphic analyses, and to Mr. Stephen 
Mikula, who prepared most of the samples 
and carried out a number of the mechanical 
tests. 

REFERENCES 

1. R. P. Heuer: Jnl. Amer. Chem. Soc. (1927) 49, 

2715. 

2. F. N. Rhines: Trans. A.I.M.E. (1940) 137, 290. 

3. A. Sieverts and W. Krumbhaar: Ztsch. physikal. 

Chem. (1910) 74, 277. 

4. P. Rontgen and F. Muller: Metallwirtschaft (iQ34) 

13, 81, 87. 

5. N. P. Allen: Jnl. Inst. Metals (1930) 43, 81. 

6. C. E. Ransley: Jnl. Inst. Metals (1939) 65, 157. 

7. W. Koster: Ztsch. Metallkunde (1928) 20, 189. 

8. W. E. Alkins: Jnl. Inst. Metals (1918) 20, 39- 

9. F. Johnson: Jnl. Inst. Metals (1919) 21, 335- 

10. W. E. Alkins: Jnl. Inst. Metals (1920) 23, 381- 

11. W. R. Webster, J. L. Christie and R. S. Pratt: 

Proc. Inst. Metals Div. A.I.M.E. (1927) 233. 



12. C. G. Maier: Trans. A.I.M.E. (1936) 122, 121. 

13. D. Stockdale: Quarterly Jnl. Inst. Metals (Aug. 

1940) 287. 

14. H. Van Bergen: Naturwiss. (1937) 25 24/25, 415. 

15. E. A. Owen and E. W. Roberts: Phil. Mag. (1939) 

-% 294- 
"ammann: Ztsch. Metallkunde (1932) 24, 221 



^ 27, 294- 
16. G. Tj 



APPENDIX 

An appendix, consisting of tables show 
ing data of Fig. 2 and annealing data, has 
been issued as auxiliary publication, and 
maybe obtained by writing to the American 
Documentation Institute, Offices of Science 
Service, 2101 Constitution Ave., Wash 
ington, D. C., asking for Document No. 1465 
and remitting 27^ for microfilm (images 
i inch high on standard 35-mm. motion 
picture film) or 90^ for photoprint (read 
able without mechanical aid). 

DISCUSSION 

(E. M. Wise presiding) 

W. M. BALDWIN, JR.,* Cleveland, Ohio. 
The question was raised during the oral discus 
sion to this paper whether there would have 
been some softening effect at room temperature 
suffered by the hard-drawn material between 
the time of preparation and time of testing. 
Some idea of the rate of softening at room 
temperature has been hinted at by the authors 
in the last three paragraphs of their paper. 
Their line of thought has here been extended. 
Fig. 6 has been constructed from the data con 
tained in their Figs. 3, 4 and 5. The times and 
temperatures of the half-hard points of the 
tensile-strength curves of wire reduced 50, 75 
and 87.5 per cent have been plotted, using the 
logarithm of the time and the reciprocal of the 
temperature. Straight lines result. Extra 
polating the lines to room temperature (2oC.) 
shows that it would take almost 9 years for the 
most severely worked samples to reach, the half- 
hard point at a room-temperature anneal and 
almost 60 years for the least severely worked 
samples to do the same. The slopes of the lines 
are seen to be the same in all three cases. 
Fitting these lines to the general equation : 

A= At E- 



* Research Metallurgist, Chase Brass and Copper 
Co. 



284 



PREPARATION AND SOME PROPERTIES OP HIGH-PURITY COPPER 



outlined by Dushman, 17 which in the logarith- the values of Qs derived from tensile-strength 

m icform vs. temperature curves (around 6000 cal.j, 

Q Dushman (loc. cit.) forwards the hypothesis 

log A = ~ -flog^o tliat in the case O f tensile-strength heats of 



TEMPERATURE - DEGREES CENTIGRADE 




RECIPROCAL 



TEMPERATURE 
FIG. 6. 



- (I/T 



1000 ) 



Q 



is of the straight-line form of 
y sx -j- b 
where s, the slope, is represented by ^> the 

values of Q (which, in this case, would be the 
energy of the annealing reaction) may be 
directly computed from the slope of the lines 
as being 22,900 calories. 

This value lies below the values of the activa 
tion energy for copper computed from self- 
diffusion data (around 60,000 cal.) and above 



S. Dushman: Proc. Amer. Soc. Test. Mat. (1929) 



reaction "the reaction involved is either a 
motion of relatively large units, such as the 
grains, or a motion of similar atoms past each 
other; the reaction in diffusion is one in which 
individual atoms of one kind have to inter 
change places with atoms of another kind. 3 
The fact that the present value of heat of 
reactions lies midway between the two values 
discussed by Dushman may indicate that the 
size of unit involved in the softening process is 
between the sizes involved in the diffusion and 
tensile strength processes. 



DISCUSSION 



285 



S. SKOWRONSKI,* Perth Amboy, N. J. The 
staff of the Research Department of the 
American Smelting and Refining Co. is to be 
congratulated; first, on the production of what 
may be called Chemically Pure Copper and 
second, on the work now being done on the 
physical properties of this super pure copper. A 
great deal of the discrepancy in the literature 
regarding the properties of copper arises from 
the fact that the experimenters used different 
types of copper with varying chemical com 
position, and with a source of pure copper now 
available we should be able to obtain more 
consistent results and better fundamental data. 

E. M. WiSE,t Bayonne, N. J. Have the 
authors studied the recrystallizing behavior of 
this extremely pure material at low reduction, 
of the order of perhaps 5 to 25 per cent? 

I ask because of some work that we did on 
platinum of very high purity. It appeared that 
the grain size of this material was determined 
almost wholly by the reduction, and the 
material was very reluctant to grow grains. 

J. CHIPMAN,} Cambridge, Mass. I want 
to report a finding that may have some bearing 
on the purity of this copper, and to ask whether 
the authors have any evidence along the same 
line. 

The porosity of copper cast in certain 
atmospheres may be explained if carbon is 
slightly soluble in liquid copper. To test this 
idea, Mr. Leonard, in our laboratories, melted 
copper in an atmosphere of nitrogen in an 
Acheson graphite crucible, pouring the copper 
through a hole in the bottom of the crucible 
directly into a tank of water. Then we analyzed 
the copper for carbon by the same combustion 
method that we use for low-carbon steel, and 
with very careful analysis we found what might 
be reported as a trace of carbon in the copper. 
The amount found did not appear to depend 
upon the melting temperature, and was always 
small; it would average 0.003 P er cent, which is 
more than the experimental errors of the 
analysis. 

Since the highly purified copper reported in 
this paper was also melted in graphite, it may 
have contained a similar proportion of carbon. 



It seems altogether possible, therefore, that the 
amount of this impurity may be three times as 
great as the sum of all the others. This would 
not necessarily be evidenced by conductivity 
data, since, as the authors point out, the solu 
bility of carbon in solid copper is probably 
exceedingly small. 

C. S. SMITH,* Waterbury, Conn. Have the 
authors noted any change in property or 
structure of their high-purity copper at room 
temperature after cold-working? Particularly, 
it would be interesting to know whether any 
change occurred during cold-drawing or in the 
interval between cold-drawing and annealing. 

Persons not connected with the copper 
industry should be cautioned that the con 
ductivities given in the paper are mass con 
ductivities based on measurement of resistance, 
length and weight, not cross-sectional area. 
The density of pure copper indeed, even of 
modern commercial copper is higher than the 
value chosen as standard in 1914, and, if 
the mass values are converted to volume 
conductivities by the use of conversion factors 
assuming the standard density of 8.89, the 
results will be low by as much as 0.74 per cent. 
This is important if comparison is to be made 
with other materials or with the results of other 
methods for the determination of conductivity. 
Except for copper conductors, conductivities 
are generally expressed in terms of dimensions 
rather than weight. In fact, the conductivity 
of the International Annealed Copper Standard 
is itself stated in terms of volume, and mass 
conductivities are used only because they are 
easier to determine. The authors figure for the 
mass conductivity of annealed pure copper 
(102.3 per cent I.A.C.S., density of 8.956 
grams per c.c.) is equivalent to a volume 
resistivity of 1.6730 microhm-cm, at 2OC., or a 
volume conductivity of 103.06 per cent I.A.C.S. 

A. M. WiGHTMAN,f North East, Pa. It is 
not proper to sharpshoot at so excellent a 
paper, but I notice that the density of a com 
mercial wirebar, with no reduction, is given as 
8.9314. Would that mean as cast? Is that not a 
rather high figure, and does it not conflict with 
such work as Corson s on the bulk density? 



* Research Chemist, Ran tan Copper Works, 
t Staff Adviser, International Nickel Co. 
t Professor of Metallurgy, Massachusetts Institute 
of Technology. 



* Research Metallurgist, American Brass Co., 
Waterbury, Conn. 

t The Electric Materials Co. 



286 



PREPARATION AND SOME PROPERTIES OF HIGH-PURITY COPPER 



C. R. HAYWARD,* Cambridge, Mass. I 
want to emphasize what has already been said 
by the chairman, that we are certainly on the 
right track in starting with very pure material. 
I have urged for many years that we pay more 
attention to minute quantities of impurities in 
copper, and I feel that now with a pure material 
to start with we may determine some of the 
real properties of this metal. 

Mr. Skowronski was one of the pioneers in 
determining the effects of various impurities in 
copper, and I know he rejoices, as all of us do 
who are interested in copper, that perhaps some 
improvement can be made in the accuracy 
of the results which at the time he made them 
were outstanding. 

In regard to Dr. Chipman s remarks, I 
wonder if his traces of carbon are not due to a 
very slight solubility of carbon monoxide, 
which obviously would be present. Contact 
with the crucible might also add some graphite, 
mechanically suspended. 

J. S. SMART, JR. (authors reply). Mr. Bald 
win s comments are welcome, as they represent 
an extension of the data along lines that the 
authors have not had occasion to pursue up 
to the present time. It is unfortunate, of 
course, that the limitations of the data neces 
sitate such a lengthy extrapolation, and the 
authors are aware that additional points would 
have been desirable. 

A comprehensive survey of the grain-size 
characteristics at low reductions has not been 
carried out, and we are not entitled to specifi 
cally answer Mr. Wise on the basis of our 
present data. However, we have observed that 
pure copper, reduced 10 to 20 per cent, exhibits 
a tendency to form a rather large recrystallized 
grain size, which is fairly stable, and prone to 
resist increase by grain growth. The initial 
grain size of the annealed strip used for the test 
deformations was rather high, and the results 
are not considered as conclusive. 

Dr. Smith s distinction between mass and 
volume conductivities is a timely contribution, 
which deserves emphasis. The possibility of the 
occurrence of a significant change in the proper 
ties of the hard-drawn wire at room tempera 
ture during the interval between drawing and 



* Professor of Process Metallurgy, Massachusetts 
Institute of Technology. 



annealing has been mathematically considered 
by Mr. Baldwin. Other general observations 
confirm the premise that detectable change did 
not occur. Following cold-drawing, a maximum 
of two hours elapsed before the annealing tests 
were begun. These were not run in any pre 
determined sequence, but entirely at random. 
The curves were progressively roughed out 
as the data accumulated, and filled in as 
necessary. Since the elapsed time between the 
beginning and end of the annealing treatments 
produced no detectable inconsistencies of the 
type under consideration in the curves, or in 
check samples run at different times, the authors 
feel that this source of error can safely be 
dismissed as negligible. 

Mr. Wightman s question affords oppor 
tunity to emphasize the fact that all of the 
present density measurements on oxygen- 
bearing copper were performed on annealed 
wire, which had been rolled from a suitable 
cast shape to rod, cold-drawn 33 per cent, and 
annealed i hr. at 6ooC. This annealed wire 
represents the o per cent reduction condition 
of Table 3. The density measurements made 
on castings are listed in columns i and 2 of 
Table 3, and these apply only to the con 
tinuously cast oxygen-free condition, with the 
indicated degree of cold-work. 

Professor Hayward and Dr. Chipman have 
both discussed the presence of some form of 
carbon as an impurity in the pure copper rod, 
a possibility that is irrefutable, but difficult to 
approach by direct means. The indirect evidence 
afforded by conductivity tests after suitable 
heat-treatment, and by internal oxidation, was 
deemed sufficient to negate the possibility of 
significant interference by soluble carbon with 
the properties of the copper. Ransley has 
demonstrated that carbon monoxide will not 
diffuse through a thin copper diaphragm, thus 
effectively eliminating solubility in this form. 
The third and most likely possibility is that of 
mechanical inclusion, and this may well 
account for the failure of Dr. Chipman s 
analyses to depend on temperature. The 
authors have no evidence that could either 
prove or disprove this circumstance, which 
would apply to any material used as a crucible, 
but feel that the slow withdrawal of the 
solidified rod from the bottom of a quiet bath 
should minimize this effect to a considerable 
degree. 



Solubility of Sulphur Dioxide in Molten Copper 



BY CARL F. FLOE* AND JOHN CETPMAN,* MEMBERS A.I.M.E. 

(New York Meeting, February 1941) 



THE system molten copper-oxygen- 
sulphur is of interest from both the 
practical and theoretical standpoints; prac 
tically, because oxygen and sulphur play 
an important role in the commercial 
production of copper, and theoretically 
because it offers a chance to study the 
reaction between an oxide gas, S0 2 , and a 
molten metal. Reactions of this type, such 
as that between CO or CO 2 and molten 
iron, are important in many metallurgical 
operations. Again, the determination of any 
gas-metal equilibrium is of theoretical 
interest and value insofar as it contributes 
to the general development of the laws for 
solution of gases in metals. This problem 
is being given an increasing amount of 
attention in modern metallurgy. 

This paper is the first of a series and 
presents the results of a determination of 
the solubility of sulphur dioxide in molten 
copper at temperatures from 1100 to 
i5ooC., and pressures from about 20 mm. 
of mercury to somewhat above atmospheric. 

REVIEW OF PREVIOUS WORK 

The solubility of gases in most liquids 
has been found to obey Henry s law, which 
states that for a fixed temperature the 
amount of gas dissolved is directly propor 
tional to the pressure. However this law 
does not hold for liquid metals. Instead, 
it appears that for many gases, particularly 



Manuscript received at the office of the Institute 
Nov. 28, 1940. Issued as T.P. 1308 in METALS TECH 
NOLOGY, April 1941- ,, ... , 

* Department of Metallurgy, Massachusetts 
Institute of Technology, Cambridge, Mass. 



the simple diatomic ones such as 2 , N 2 and 
H 2 , the amount dissolved in liquid metals is 
directly proportional to the square root of 
the pressure. This is known as Sieverts 5 law. 
It may be recognized as a special case of 
Henry s law, which applies when the solute 
gas is dissociated into atoms or when it 
reacts with the solvent metal to form 
solute molecules containing single atoms 
of the gaseous elements, as, for example: 



[i] 



M*X. 



which, if the solubility is not too high and 
the concentration of metal therefore nearly 
constant, can be simplified to: 



MG 



Following the same reasoning, it would 
be expected that the solubility of S0 2 in 
copper would be proportional to the cube 
root of the pressure, the reaction being as 
follows: 



6Cu -f S0 2 



Cu 2 S + 
Cu 2 S X 



[3] 



or, neglecting the change in metal con 
centration and considering that the con 
centrations of sulphur and oxygen are both 
proportional to the volume V of gas 
absorbed: 



or 



> [5] 



287 



2 88 



SOLUBILITY OF SULPHUR DIOXIDE IN MOLTEN COPPER 



The only very extensive determination 
of the solubility of sulphur dioxide in 
copper was made by Sieverts and Krumb- 
harr. 1 Contrary to expectation, the solu 
bility was found to be more nearly 
proportional to the square root of the 
pressure than to the cube root, which could 
not be explained satisfactorily. Smithells 2 
made a plot of the cube root of the pressure 
against the amount of gas dissolved in 
Sieverts experiments and concluded that 
the experimental points lay very nearly on 
straight lines, which, however, did not pass 
through the origin. 

Stubbs, 3 on the basis of Sieverts data, 
proposed that in this case the total solu 
bility was a combination of molecular solu 
tion and chemical reaction, the former 
being proportional to the first power of the 
pressure and the latter to the cube root of 
the pressure as follows: 



k P + k" \/P 



[6] 



By calculation he found that at atmospheric 
pressure about 70 per cent of the total gas 
dissolved reacted chemically to form Cu 2 S 
and Cu 2 O while about 30 per cent was 
dissolved molecularly. At lower pressures 
the proportion chemically absorbed was 
considered to be greater, at higher pres 
sures less. It is interesting to note that the 
method could not be successfully applied to 
the one experimental point available below 
100 mm. pressure. Reference will be made 
to this later in connection with the present 
results. 

The theory of molecular absorption 
does not seem reasonable in view of the 
great size of the S0 2 molecule. The more 
recent proposal by Lepp 4 that solution of 
all gases in metals is a chemical process 
that can be explained on the basis of 
thermodynamics and chemical equilibrium 
seems to be more acceptable. However, 
Lepp explains that for the case of sulphur 
dioxide in copper the normal cube-root 



1 References are at the end of the paper. 



relationship that would be expected from 
equations 4 and 5 does not hold because 
at the lower temperatures the dissociation 
of the sulphide phase is negligible and 
therefore the limit of solubility is deter 
mined by the dissociation of Cu 2 0. The 
absorption then becomes proportional to 
the square root of the pressure. At higher 
temperatures the dissociation of Cu 2 S 
becomes appreciable and the root of the 
pressure factor increases toward a maxi 
mum of 3. This reasoning is difficult to 
follow, as it would seem that such tend 
encies for dissociation are taken fully 
into account in the determination of the 
ordinary equilibrium constant for the 
reaction. 

Sieverts and Bergner 5 conducted further 
experiments on the solubility of sulphur 
dioxide in copper in which an excess of 
oxygen or sulphur was present. The results, 
however, did not make it possible to draw 
any further conclusions as to the nature 
of solution. An excellent review of the 
general subject of gases in molten copper 
has been made by Ellis. 6 After reviewing 
the important work he concludes that "the 
solubility in copper of sulphur dioxide per 
se has yet to be determined." 

In general it may be concluded that there 
is much disagreement as to the nature of 
the equilibrium between sulphur dioxide 
and molten copper. Since only one set of 
data, determined 30 years ago, is available, 
it was decided that a redetermination of the 
solubility limits would be of considerable 
value. 

EXPERIMENTAL METHODS 

The apparatus used for the determina 
tion of solubility, which in principle is 
similar to that first used by Sieverts, is 
shown schematically in Fig. i. In making a 
run, 100 grams of copper was first melted 
in vacuum in the fused silica tube A using 
induction heating from the copper coil B. 
Induction heating was used principally 
because the stirring action on the molten 



CARL F. FLOE AND JOHN CHIPMAN 



289 



metal permitted equilibrium to be estab 
lished more rapidly. Temperatures were 
read by a platinum-platinum rhodium 
thermocouple in a small silica tube T, 



the first quantity of gas admitted had been 
reached, the stopcock D was again opened, 
another measured quantity of gas admitted, 
and the next equilibrium pressure deter- 




FIG. i. APPARATUS USED FOR DETERMINING SOLUBILITY OF SULPHUR DIOXIDE IN COPPER. 



which was connected to an accurate 
potentiometer. The melting point of pure 
copper was used as a reference. Hand regu 
lation of the power input permitted temper 
ature control within 2C. 

After the copper was melted, the first 
equilibrium point was determined by 
admitting a measured amount of SOs gas 
from the burette C into the furnace side 
of the system through the stopcock D. 
Absorption of this gas by the molten copper 
continued until a constant pressure was 
reached. This usually took from 30 min. to 
i hr. The equilibrium pressure was read 
either by the closed-tube mercury manom 
eter E or the open-tube mercury manom 
eter F. After the equilibrium pressure for 



mined. This was repeated until the pressure 
was somewhat above atmospheric. 

In most runs two temperatures, 1100 
and i2ooC. were used. In order to check 
the reversibility of the reaction, in some 
cases a point was first determined at 
nooC. then the temperature was in 
creased to cause more absorption, and 
finally reduced again to iiooC. in order 
that equilibrium might be reached by 
evolution of the gas. Above 100 mm. pres 
sure the two results usually agreed within 
less than 0.5 per cent. However, below 
100 mm. the difference was sometimes as 
much as 15 per cent. In general, results for 
different runs at low pressures were also 
inclined to be erratic. Pressures below 10 



290 



SOLUBILITY OF SULPHUR DIOXIDE IN MOLTEN COPPER 



mm. could not be determined with very 
great accuracy. 

Calculation of the composition of the 
molten copper in contact with the S0 2 gas 
at a given equilibrium pressure involves 
knowing how much of the total gas ad 
mitted to the furnace side of the system has 
been absorbed. This can be calculated if 
the so-called "hot volume" is known, or 
the volume of gas at standard conditions 
necessary to fill the furnace side of the 
system after the copper has been heated 
to the temperature at which the solubility 
is to be determined. This volume, of course, 
is less than the actual volume because of 
expansion of the gas in the vicinity of the 
molten copper. The determination of the 
hot volume was carried out with purified 
nitrogen, which is virtually insoluble in 
liquid copper. 

The accuracy of solubility determinations 
made by this method is increased if the hot 
volume is kept as small as possible. In this 
work, owing to the design of the melting 
chamber and to the use of capillary tubing 
for all connections on the furnace side of 
the system, it was never greater than about 
30 c.c. 

For measuring gas into the furnace sys 
tem a fixed mercury leveling bottle, G, was 
used, operating from the vacuum system. 
This permitted the mercury level in the 
burette to be controlled simply by the 
opening or closing of the stopcocks H and /. 
The sulphur dioxide from the cylinder J 
was dried by passing through phosphorus 
pentoxide before admission to the burette. 

Commercial oxygen-free copper, O.F.- 
H.C. brand, in the form of Jg-in. wire, was 
used for all experiments. This was kindly 
furnished by the Anaconda Copper Co. A 
check analysis after melting in vacuum in 
the furnace showed both the oxygen and 
the sulphur to be less than o.ooi per cent. 

RESULTS or EXPERIMENTAL WORK 

The results for solubility determinations 
at temperatures ranging from 1100 to 
1 5ooC. and varying pressures are given in 



Table r and graphically in Figs. 2 and 3. 
The last column of Table i shows the calcu 
lated concentration of sulphur or oxygen in 
molten copper that is in equilibrium with 
the sulphur dioxide gas at the various pres 
sures given. 

TABLE i . Solubility of Sulphur Dioxide in 

Molten Copper at Various Temperatures 

and Pressures 



Run 
No 


Temper 
ature, 
Deg. C. 


Equilib 
rium 
Pressure, 
Mm. Hg 


S0 2 Ab 
sorbed per 
100 Grams 
Cu, C.c. 


Sulphur or 
Oxygen, 
Per Cent 


6 


1 100 


13-9 
20.7 
30.9 
56.1 
129.0 
642.0 
893-0 


8.3 
19.9 
29.2 
44-5 
77-6 
146.4 
165.5 


0.0119 
0.0284 
0.0417 
0.0636 

O. Ill 

0.209 
0.236 


1 200 


19.6 
28.0 
43-9 

100. 

410.0 

583.5 
796.0 


19-1 
29-3 
44-9 
78.4 
155-3 
179-0 
204.4 


0.0271 
0.0419 
o . 0642 

O.II2 
O.222 
0.256 
0.292 


7 


1 100 


IT. 6 
37-0 
89.5 
309.5 
604.4 
898.5 


20.6 

45.9 
69.4 

no. 4 

144-5 
170.5 


0.0294 
0.0656 
0.0991 
0.158 
0.206 
0.244 


1200 


9.8 
25.1 
56.0 
187.0 
371.0 
590.0 
804.0 


20.7 
46.2 
70.4 
114-8 
153-6 
184.5 
208.5 


O.0296 

o . 0660 

0.1005 
0.164 
O.220 
0.264 
0.298 


1300 


752.0 


251.0 


0-358 


8 


IIOO 


13.8 
50.0 
145-0 
376.0 
654-0 
914-0 


19.8 
SO.i 
8o.O 
116.4 
146.9 
167.8 


0.0283 
O.07I6 
O.II4 
0.166 
0.210 
0.240 


1200 


12.0 

34-0 
90.O 
231.0 
420.0 
614.0 
830.0 


19.8 
50.6 

81.7 

121. 7 

156.8 
182.3 
203.6 


0.0283 
O.O723 

O.II? 
0.174 
0.224 
0.260 
0.290 


1300 


711.0 


242. 


0.346 


9 


1200 


771.0 


197. 


0.281 


II 


1400 
1500 
1200 


755-0 
755.0 
760.0 


300. 

352. 

211. 


0.428 
0,501 
0.301 


12 


1500 


764.0 


365. 


0.522 



Fig. 2 graphically illustrates how the 
volume of sulphur dioxide absorbed per 100 
grams of copper changes with pressure at 



CARL E. FLOE AND JOHN CHIPMAN 



SQI 



1100 and i2ooC. The points obtained for reaction between the fused silica tube and 

the different runs, No. 6, No. 7, and No. 8, the melt, the amount of this reaction vary- 

are shown by different symbols. These ing with the time and temperature. Another 

points show that each set lies on a fairly reason is possible slight differences in the 



200 



160 



g.120 



fc 



801 



O 

u 



40 




s- x e 

Sieverfi H23f*C.= 



W 6" 200 400 600 800 

Pressure mm. Hj 

FIG. 2. SOLUBILITY OF SULPHUR DIOXIDE IN COPPER AS A FUNCTION OF PRESSURE. 

400i 1 1 1 1 1 



U300 
E 



&200 



O 100 



d 
O 



1000 



1100 



1400 



1500 



J600 



1200 1300 

Temp. C. 

FIG. 3. SOLUBILITY OF SULPHUR DIOXIDE IN COPPER AT ONE ATMOSPHERE PRESSURE AS A FUNC 
TION OF TEMPERATURE. 



smooth curve, but that the results for differ 
ent runs do not duplicate one another 
exactly, although they are reasonably close 
at the higher pressures. One reason for lack 
of duplication is that there is a slight 



composition of the original copper, although 
this effect is not believed to be large. Runs 
No. 8, No. 9, No. 11, and No. 12 were deox 
idized with hydrogen and then evacuated 
before any sulphur dioxide gas was admitted 



SOLUBILITY OF SULPHUR DIOXIDE IN MOLTEN COPPER 



to the furnace system, while runs No. 6 
and No. 7 were not hydrogen-treated. 

Fig. 2 shows also the results obtained by 
Sieverts for a temperature of ii23C. In 
general, the solubilities found are somewhat 



where F S02 is the number of cubic centi 
meters of sulphur dioxide at one atmos 
phere dissolved per 100 grams of copper, 
and t is the temperature in degrees 
centigrade. 



200 



160 



o 
2 120 



80 



O 40 




FIG. 4, SOLUBILITY OF SULPHUR DIOXIDE IN COPPER AS A FUNCTION OF THE CUBE ROOT OF THE 

PRESSURE. 
See Fig. 2 for an explanation of the symbols. 



higher than those obtained by Sieverts, 
although the general shape of the curve is 
the same. The tendency for somewhat 
erratic results at low pressures was not 
noted by Sieverts because very few points 
were determined in this region. Fig. 2 shows 
also that at pressures below 20 mm. the 
temperature coefficient of solubility was 
very small. With increasing pressure, this 
coefficient increased rapidly. 

Fig. 3 shows the way in which the solu 
bility of sulphur dioxide in molten copper 
at one atmosphere pressure varies with 
temperature from the melting point to 
i5ooC, This solubility is directly propor 
tional to the temperature. The equation 
for the curve within the limits of error of 
the experiments may be expressed by: 

Vao* 400 + 0.50^ [7] 



There was considerable evolution of 
sulphur dioxide gas as the copper solidified. 
The amount of this could not be measured 
accurately because of cracking of the silica 

TABLE 2. Analysis of Ingots for Sulphur 
and Oxygen 









Cubic Centi 


Cubic Centi 


Run 
No. 


S, Per 
Cent 


0, Per 

Cent 


meters Re 
maining 
from S 


meters Re 
maining 
from O 








Analysis 


Analysis 


6 


0.081 




57 




7 


0.071 




50 




8 


0.047 


0.073 


33 


5i 


9 


0.039 


0.044 


27 


3i 


II 


0.032 


0.050 


22 


35 



tube as solidification took place. The best 
estimate that could be made indicated that 
the amount remaining in the solid ingot 
was of the order of 50 c.c. per 100 grams of 



CARL F. FLOE AND JOHN CHIPMAN 



2 93 



copper. This is present as cuprous oxide, 
cuprous sulphide, and occluded gas in the 
blowholes formed during freezing. All of 
the ingots from these melts showed high 
porosity. 



coordinate and the volume of gas dissolved 
as the other. This is shown in Fig. 4. The 
points are represented reason ably well by a 
straight line at both 1100 and i2ooC., 
but the extension of these lines would in 



200 



160 



120 



*0 



40 




15 



20 



FIG. 5. SOLUBILITY OF SULPHUR DIOXIDE IN COPPER AS A FUNCTION OF THE SQUARE ROOT OF 

THE PRESSURE. 
See Fig. 2 for an explanation of the symbols. 



The chemically combined sulphur and 
oxygen present in the ingots after solidifi 
cation was determined by analysis, with 
the results shown in Table 2. 

The oxygen was determined by the 
vacuum-fusion method, the accuracy of 
which for materials of this type is question 
able. The results for sulphur show that the 
amount retained after solidification varies 
considerably, being equivalent to from 22 
to 57 c.c. of sulphur dioxide. 

DISCUSSION OF RESULTS 

When Sieverts law is extended to tri- 
atomic gases such as 862, it would be ex 
pected that the volume of gas dissolved 
would be approximately proportional to 
the cube root of the pressure. The results 
given in Table i therefore have been plotted 
using the cube root of pressure as one 



both cases intersect the pressure axis at a 
value of about 1.6 rather than at the origin. 
Smithells 2 has shown that Sieverts results 
give a similar curve. 

The solubility above 20 mm. may be 
expressed by an equation of the form: 

F S02 = K(\/P - 1.6) [8] 

for iiooC., K = 21.0, while for i2ooC., 
K = 27.0. However, it is believed that this 
equation is only of empirical significance, 
since there is no obvious explanation for it 
in terms of any reasonable hypothesis re 
garding the nature of the solution. 

Sieverts suggested that his results more 
nearly showed the solubility to be propor 
tional to the square root of the pressure. 
To check this, the values given in Table i 
were plotted in this way, as shown in 
Fig. 5. This was not as successful as plotting 



294 



SOLUBILITY OF SULPHUR DIOXIDE IN MOLTEN COPPER 



the cube root of the pressure. The extension 
of the nearly straight-line portion of the 
curves intersects the vertical axis above 
the origin, while the values for low pres- 



expected from the log plot; namely, that 
at pressures above about 100 mm. a straight 
line is obtained, the extension of which 
would pass through the origin. However, 



1.5 



i.o - 




1.0 1.5 2.0 2.5 3.0 

Lofi. P 

FIG. 6. LOGARITHMIC PLOT OP NUMBER OF CUBIC CENTIMETERS OF SULPHUR DIOXIDE DISSOLVED 

PER 100 GRAMS OF COPPER VERSUS LOGARITHM OF PRESSURE. 

See Fig. 2 for an explanation of the symbols. 



sures fall off rapidly and, if extended, would 
intersect the pressure axis at about 3. 
There is, of course, no fundamental reason 
why the amount of sulphur dioxide 
absorbed by molten copper should be pro 
portional to the square root of the pressure. 

Next, in order to try to interpret 
the results, a plot of the logarithm of 
the amount of SOz absorbed versus the 
logarithm of the pressure was made for 
runs No. 6, No. 7, and No. 8 at 1100 and 
i2ooC (Fig. 6). All of these curves 
approach almost the same constant value 
of slope at high pressures, the average 
being 2,46. However, as the pressure is 
decreased, the slope rapidly decreases, 
indicating that it is not possible to make 
the absorption proportional to any single 
root of the pressure over the entire range. 

In order to show the effect of raising the 
pressure to the 1/2.5 power, a graph was 
made (Fig. 7). The result is what would be 



at low pressures the values again fall away 
toward the pressure axis. 

These results are interpreted as indi 
cating that no simple reaction between the 
gas and the metal can account for the data. 
The failure of the initially postulated 
reaction (Eq. 3) is demonstrated graphi 
cally in Fig. 8, where the ratio 



[9] 



is plotted against the pressure. It is evident 
that the "equilibrium constant" is never 
strictly constant and actually appears to 
approach zero at low pressures! In these 
calculations, the concentration of copper 
has been assumed to be constant. If the 
proper values for copper concentration were 
used, it would shift the results only slightly, 
but in such a way as to increase the vari 
ability of the ratio. 



CARL F. FLOE AND JOHN CHIPMAN 



295 



Stubbs 3 attempted to explain these dis 
crepancies by assuming that a part of the 



for run No. 8 at i2ooC. The results are 
shown in Table 3. The amount of assumed 



sulphur dioxide was dissolved molecularly chemical solubility was 50 per cent, 70 per 
as shown by Eq. 6. This can be done fairly cent and 90 per cent, respectively, of the 




JT IG> 



y w SOLUBILITY OF SULPHUR DIOXIDE IN COPPER AS A FUNCTION OF THE 1/2.5 POWER OF 

PRESSURE. 
For an explanation of the symbols, see Fig. 2. 




FIG. 8. " EQUILIBRIUM CONSTANT" FOR REACTION BETWEEN SULPHUR DIOXIDE AND COPPER 
(EQUATION 3) AS A FUNCTION OF PRESSURE. 
See Fig. 2 for an explanation of the symbols. 

successfully for pressures above 100 mm., total sulphur dioxide dissolved at the 

but becomes unreasonable when the pres- highest pressure. These results show that 

sure is below 100 mm. To iUustrate this, for pressures below 100 mm. the constant 

values of K" from Eq. 6 were calculated is far too smaU, which agrees with the 



296 



SOLUBILITY OF SULPHUR DIOXIDE IN MOLTEN COPPER 



TABLE 3. Values of K" of Equation 6 
Calculated from Assumed Partial Molecular 

Solution 
RUN No. 8, i2ooC. 



Per cent SOa chemically dis 
solved at 830 mm .... .... 


& 

Cent 


70 
Per 
Cent 


P 9 e r 
Cent 




Corresponding values for K f . . 


0.1226 


0.0736 


0.0246 


Values for K" at P 
12 mm . .... ... . , 


8.0 
14.3 
15.8 
15.2 
14.1 

12.6 

ro.8 


8.3 
14.8 
16.8 
17. i 
16.8 
16.1 
15-2 


8-5 
15-4 
17-7 
18.9 
19.6 
19.7 
19.5 


34 mm 


90 mm . . . .... 


231 mm 


420 mm 


614 mm . 


830 mm . . 





data plotted in Fig. 8. No assumption of 
partial molecular solubility can correct this 
discrepancy. 

In all of these calculations for equilib 
rium constants it was assumed that the 
activity of the sulphide and oxide dissolved 
in the copper was proportional to the 
concentration. This would be true for 
perfect solutions, but may not be true over 
the range of compositions covered. The 
conventional interpretation would be that 
activity is proportional to concentration in 
dilute solutions, but that the ratio of 
activity to concentration decreases with 
increasing concentration. For the data 
given in Table 3, this would imply that 
the more concentrated solutions deviate 
widely from ideal behavior, a conclusion 
that is contradicted by the observed normal 
lowering of the melting point of copper 
both by the oxide and by the sulphide. 7 

These considerations lead to the tenta 
tive conclusion that the anomalous results 
are due to irregular behavior of the dilute 
solutions rather than to deviations from 
ideality at the higher concentrations. 

In attempting to explain the low solu 
bility at low pressures, it was first thought 
possible that the sulphur dioxide was 
broken down to sulphur monoxide by 
solution of one atom of the oxygen. This 
would have the effect of causing the pres 



sure to remain constant or at a higher than 
normal value. However, calculations based 
on the data for the free energies of sulphur 
monoxide and sulphur dioxide given by 
Montgomery and Kassel 8 and Cross 9 show 
that this effect would be negligibly small. 

SUMMARY AND CONCLUSIONS 

The solubility of sulphur dioxide in 
molten copper has been determined over 
the range of pressure from about 20 mm. 
Hg to 900 mm. Hg, and temperatures from 
1100 to i5ooC. The data show that the 
solubility cannot be interpreted by postu 
lating any simple chemical reaction between 
the gas and the metal. The equation for 
solution that would normally be written 
(Eq. 3) does not hold, particularly at low 
pressures. The indications are that this is 
due to some irregular behavior of the 
dilute solutions, which cannot be readily 
explained on the basis of our present knowl 
edge of the nature of solution of gases in 
molten metals. 

REFERENCES 

1. Sieverts and Krumbharr: Ztsch. phys. Chem. (1910) 

2. Smithells: Gases and Metals, 158. London, Chap 

man and Hall, Ltd. 1937. 

3. Stubbs: Jnl. Chem. Soc. (London) (1913) 103, 

1445 

4. Lepp: Bull. Assoc. Tech. Fonderie (1937) n, 84. 

5. Sieverts and Bergner: Ztsch. pkys. Chem. (1913) 

82, 257. 

6. Ellis: Trans. A.I.M.E. (1933.) 106, 487. 

7. Hansen: Aufbau der Zweistofflegierungen, 589, 

6 10. Berlin, 1936. Julius Springer. 

8. Montgomery and Kassel: Jnl. Chem. Phys. (1934) 

2, 417. 
9- Cross: Jnl. Chem. Phys. (1935) 3, 825. 

DISCUSSION 

(5. Skowronski presiding) 

0. W. ELLIS,* Toronto, Ont. There is an 
interesting point in connection with the method 
of melting the copper described by the authors. 
The question arises whether the temperature 
of the gas immediately above the metal and 
somewhat distant from the metal is at the same 
temperature as the metal itself. 

Some years ago the writer propounded the 
following question to Prof. N. Rashevsky: 

* Director, Department of Engineering and Metal 
lurgy, Ontario Research Foundation. 



DISCUSSION 



297 



Suppose a hot body maintained at a tempera 
ture TK is put into an atmosphere of any gas 
having a pressure P and a temperature T . 
When a stationary state is reached, the tem 
perature of the gas in the immediate neighbor 
hood of the body will be T, while at a point far 
enough remote it will be T , The variation of 
temperature from point to point will depend 
on the shape and size of the body, as well as 
on the heat conductivity of the gas. At first 
glance it might appear that the pressure within 
the gas would be the same at all points, since 
the gas is supposed to be at rest and no con 
vection currents are assumed. However, more 
detailed consideration shows that this is not 
true. The question arises whether the pressure 
of the gas will be the same at the surface of the 
body as at some distance from it; that is, 
whether it will be equal at all points to P or 
not. Professor Rashevesky s solution of the 
problem follows: 

"A rigorous solution of the problem is, 
however, very difficult since we have here to 
deal with a nonequilibrium state and the rela 
tions of the kinetic theory, which are all 
obtained on the assumption of a thermal 
equilibrium that is, of a temperature constant 
throughout the whole gas do not apply here. 
The problem is intimately connected with the 
theory of radiometric forces, and up to the 
present no entirely satisfactory solution of 
these problems has been given. The earlier 
works of M. Knudsen dealt principally with 
gases at low pressures. Without this restriction 
the problem has been investigated by West- 
phal, but the solution was later on recognized 
as incorrect by the author himself. A paper by 
E. Einstein [Ann. Phys. (1922) 69, 241] treats 
the problem in a very general way, but the 
final formulas are of no use for the case which 
interests us here. A simplified treatment of the 
problem has been given more recently by 
E. Einstein [Ztsch. Phys. (1924) 27, i] in which 
all assumptions are schematical, and only the 
order of magnitude of the results, not an exact 
theory, is claimed. 

"Following here the same simplified assump 
tions as in A. Einstein s paper, we will gen 
eralize them so as to obtain an answer to the 
question that interests us. 

"We first of all suppose that the hot body is 
limited by a plane, which we choose as the yz 
plane of our system of coordinates. Then there 



is a drop of temperature along the x axis, 
which drop is supposed to be given at every 
point. If we neglect, as does Einstein, the 
velocity distribution of molecules and assume 
that they all have the same velocity u which is 
connected with the temperature T of the gas 
by the relation 

where m is the mass of the molecule, and K is 
Boltzmann s constant, then through a unit 
area perpendicular to the x axis }^nu molecules 
will pass per second, where n is the number of 
molecules in a cubic centimeter. But since there 
is no transport of gas, the number of molecules 
passing the area considered in the positive 
direction of x must be equal to the number of 
molecules passing in the negative direction. 
Also, there should be no accumulation of gas in 
certain regions. Hence mi must be constant, 
which together with Eq. i shows that n is 
inversely proportional to -\/T. More exactly 
this is seen from the following: 

"Consider a general case when there is a 
transport of mass of gas due to inequality of 
temperatures, as well as to inequality of con 
centrations. The case of inequality of concen 
tration alone, at uniform temperature, would 
lead to the usual phenomenon of diffusion. 
In the more general case we proceed in a way 
similar to that for diffusion. Consider two 
planes parallel to yz } namely, x and x + A:r. 
The number of molecules contained in a 
parallelepiped formed by two unit areas on x 
and x + A# is nkx. The variation of this num 
ber per second is equal to the difference of the 
number of molecules entering through the plane 
.T, and that leaving the plane x + Ax. The 
first is 

%ftxUx [2] 

where the index x shows that the corresponding 
quantities are taken at the point x. The second 
is equal to 

dltx&x 



- (n x 



) x -f AJC 



- f n x u x 



-f n x 






neglecting quantities of higher order. 
"Hence the difference equals 

dn x dn x \ 



298 



SOLUBILITY OF SULPHUR DIOXIDE IN MOLTEN COPPER 



and this should be equal to 

dn , 



dn\ dn 



Hence 



For a stationary state, such as we consider 

dn 

- o, and Eq. 3 reduces to 



du dn d(nu) 



Hence 



nu = constant, or using Eq. i 



VjC J "T 1 
- - 
m 



and 



constant 



const 



[si 



[6] 



"If the body is at a temperature of i2ooK., 
while the surrounding gas (say air) is at 3ooK., 
the concentration in the immediate neighbor 
hood of the boundary of the body will be two 
times smaller than at a remote point. But with 
the approximations used here, there is also 
for a nonuniform temperature in the gas 

P = nKT [7] 

Combining this with Eq. 6 gives 



P = B 



= AK) 



[8] 



" Hence, the pressure increases as the square 
root of temperature. In the case just mentioned, 
the pressure will be twice as large at the bound 
ary as at a remote point. 

"With the degree of approximation used 
here, we may apply these formulas also to 
partial pressures and concentrations, say of 
HsO vapors in air. 

"It should be remarked that although both 
pressure and concentration in the neighborhood 
of the hot body differ from those at a remote 
point, the number of molecules striking a unit 
surface of the body per second is the same as 
though the body were cold. In applying these 
considerations to the calculations of solubility 
of water vapor in a melted metal exposed to 
air, by applying Henry s law, one has therefore 
to consider the question as to what is character 



istic for the solubility of a gas, the pressure, 
or the quantity of molecules striking the surface 
of the metal per second. 

"As mentioned above, the considerations 
here give only an estimation of the order of 
magnitude of the quantities involved." 

It will be observed that Professor Rashevsky 
refers in particular to "calculations of solu 
bility of water vapor in a melted metal exposed 
to air." This was the particukr gas with which 
the writer was concerned at the time of his 
inquiry. The writer feels, however, that 
whether water vapor or sulphur dioxide is in 
question, some consideration should be given 
to Professor Rashevsky s answer, in view of 
its quite general application. 

In other words, the writer feels it desirable to 
ensure, if possible, that the temperature of the 
gas above the molten metal be held throughout 
at the same temperature as the molten metal 
itself. This, of course, would involve a change 
in the design of the furnace used in solubility 
experiments. 

S. SKOWRONSKI,* Perth Amboy, N. J. 
Formerly when discussing the solubility of 
sulphur dioxide in copper we had to refer to 
Sieberts work, but technically this work is 
ancient and we are indebted to the authors of 
this paper for furnishing us with more up-to- 
date data, more in line with the present 
theories of gases in metals. 

Commercially there is another phase of the 
problem that should be investigated; that is, 
the effect of sulphur dioxide in copper in the 
presence of hydrogen, with a diminishing 
amount of oxygen in the copper. There is no 
question in my mind that at some stage hydro 
gen sulphide must be formed, but this has 
never been proved, and if it could be proved 
it would go a long way toward explaining the 
overpoling phenomenon and the peculiar 
behavior of gassy copper. 

As far back as 1918, Carpenter and Elam, in 
their classical work on the gases in Admiralty 
bronze (88-10-2), found that the predominating 
gas was hydrogen sulphide; they also found 
indications of this gas in one brand of refined 
copper, and reported hydrogen sulphide or 
sulphur dioxide as the predominating gas in 
this brand of copper. Research work is needed 

* Research Chemist, Raritan Copper Works. 



DISCUSSION 



299 



to differentiate between the action of hydrogen 
sulphide and sulphur dioxide in refined copper. 

C, R. HAYWARD.* Cambridge, Mass. In 
some work done by myself and two co- 
workers, 10 perhaps the most astonishing thing 
found was the behavior of sulphur dioxide as 
it came in contact with copper flowing from 
a crucible to a mold. 

The other gases hydrogen, nitrogen and 
various others were very mild in their effect, 
but sulphur dioxide would attack the copper 
with great avidity. A crucible containing 
about 30 Ib. of molten copper at 1150 was 
opened with a plug at the bottom, and the 
copper flowed into a mold into which a small 
stream of liquid sulphur dioxide was flowing. 
The copper absorbed the gas and swelled to 
nearly twice its volume, and the odor of 
sulphur dioxide on the copper after it was taken 
out was noticeable. This copper definitely 
showed considerable oxygen, although copper 
melted in the same way poured into nitrogen 
showed no sign of oxygen. 

We do not know how complete the reaction 
was and to what extent sulphur dioxide was 
dissolved in the copper, but there was con 
siderable indication that sulphur dioxide gas 
was taken up in that very short period of time 
and evolved subsequently, giving the very 
porous ingots that were obtained. 

Of course, the commercial possibilities belong 
in the part of the curve that is most irregular, 
as far as the results obtained by the present 
authors are concerned. 

When Mr. Skowronski suggests the possi 
bility of the presence of hydrogen sulphide 
I do not believe that we can ignore him. I 
believe that there is a possibility that hydrogen 
sulphide may be one of the gases involved, 
and since the low pressures most interest the 
commercial copper refiner, I hope that someone 
will more carefully study this portion of the 
curve and in so doing consider the suggestions 
made by the chairman. 

Also, in further work on this subject, I hope 
that some sulphur-free copper can be used as 
a basis of the experiments. 

J. CHIPMAN. I want especially to thank 
Mr. Skowronski for his suggestion regarding 

* Professor of Process Metallurgy, Massachusetts 
Institute of Technology. . 

i" C. R. Hay ward: Effect of Atmosphere during 
Casting of Copper. Jnl. Inst. of Metals. 



the study of hydrogen sulphide in liquid copper. 
One of our major objects in presenting this 
work at this stage was that we hoped to get 
just such suggestions, which will be valuable to 
us, I am sure, in future work along this line. 
The very strong avidity of liquid copper for 
sulphur dioxide, which Professor Hayward 
mentioned, is of course obvious in the solubility 
data when you think of a hundred grams of 
copper, which has a volume of about 12 c.c., 
and which can dissolve an amount of gas that 
would occupy 200 to 300 c.c. at standard con 
ditions and actually occupies perhaps 1000 to 
1500 c.c. at the temperature of liquid copper. 
That is a reaction that obviously has con 
siderable avidity. It is one of the highest gas- 
solubility phenomena that we know of. 

C. F. FLOE AND J. CHIPMAN (authors 
reply). The question of the effect of using 
induction heating, raised by Dr. Ellis, was 
considered. It was concluded that some tem 
perature difference between the gas above the 
metal and the metal itself would exist, but that 
this would not affect the solubility. Since 
Dr. Ellis comments were received, the question 
has again been considered. Several experts in 
kinetic theory have been consulted, particularly 
Dr. Walter H. Stockmayer, of the Department 
of Chemistry, Massachusetts Institute of 
Technology. The conclusion is that the calcula 
tions quoted by Dr. Ellis do not apply to the 
apparatus used in these experiments. Kinetic 
theory and experiment concur in the prediction 
that as long as the mean free path of the 
molecules is small compared to the diameter 
of the tube no pressure differences can persist, 
even though the temperature be different 
in different sections of the tube. The equality 
of pressure is maintained through mass flow 
of the gas (convection) . Rashevsky postulates, 
incorrectly, the absence of convection. 

The phenomena discussed by Rashevsky are, 
however, predicted by kinetic theory for the 
case in which the tube diameter is much smaller 
than the mean free path. For such cases the 
constancy of p/\ff is predicted, and was 
experimentally verified as far back as 1879. u 
A brief discussion of this phenomenon, known 
as thermal transpiration, is given by Kennard. 12 



u Reynolds: Phil. Trans. (1879) 170 (II), 727. 
12 Kennard: The Kinetic Theory of Gases, 66-67. 
New York, 1938. McGraw Hill Book Co. 



300 



SOLUBILITY OP SULPHUR DIOXIDE IN MOLTEN COPPER 



The calculations of A. Einstein and E. Ein 
stein quoted apply only to the case where the 
pressures are low and the vessel dimensions are 
small compared with the mean free path of the 
molecules. In the present case it can readily 
be shown that the mean free path is small 
compared with the tube diameter. To calculate 
this, use may be made of the formula: 



kT 



L = 



where <r is the molecular diameter of SOo, found 
to be approximately 3.4 A. from viscosity data. 
Taking the worst case of lowest pressure 
(10 mm.) and highest temperature (i8ooK.) 
it is found that L ^ 0.0036 cm., which is small 
compared with the tube diameter used in the 
experiments. The conclusion is, therefore, that 
the pressures read on the manometer are equal 
to the pressures of SO* above the liquid copper. 
Dr. Ellis, on the basis that some thermal 
transpiration takes place, concludes that the 
actual pressure at the liquid gas interface 
would be greater than at a remote point. This 
is interpreted as meaning that the actual 
pressure at the boundary between the liquid 
metal and the gas would be greater than the 



pressure measured on the barometer. Such a 
condition, if anything, should have the effect 
of increasing the gas solubility, whereas the 
data actually show that, at low pressures 
particularly, the solubility is less than is 
predicted by Sieverts law. 

Professor Hayward comments that from 
the commercial point of view one is most 
interested in the irregular part of the curve. 
This is true and experiments to determine the 
very low-pressure solubilities are under way. 

The authors believe that many of the 
discrepancies in the low-pressure region are 
due to small amounts of impurities in the 
metal. Experiments now under way show that 
an excess of either oxygen or sulphur in the 
metal greatly reduces the solubility of sulphur 
dioxide. Probably the maximum in solubility 
will occur when the initial copper is of highest 
purity. It is the authors feeling that dis 
crepancies from one run to the next, and also 
discrepancies in the low-pressure region, are 
principally due to the presence of small 
amounts of impurities in the original material. 

Plans for further work include studies on the 
effect of impurities and the development of 
apparatus that will give more precise results at 
low pressures. 



Solubility of Oxygen in High-purity Copper 

BY ARTHUR PHILLIPS,* MEMBER, AND E. N. SKINNER, JR.,J JUNIOR MEMBER A.I.M.E. 

(New York Meeting, February 1941) 



DURING the course of an experimental 
program concerned with the hydrogen 
embrittlement of copper containing oxygen 
in concentrations within the solubility 
limits it became necessary to make a critical 
appraisal of the literature dealing with the 
solubility relationships of oxygen and 
solid copper. The latest determinations, 
and perhaps the only data obtained on 
"spectroscopically" pure copper, are those 
of Rhines and Mathewson 1 who reported 
that the solid solubility increases from 
0.007 per cent oxygen at 600 C. to about 
0.015 per cent at io5oC. Hanson, Marryat 
and Ford 2 had previously placed the 
solubility limit at less than 0.009 per cent 
oxygen at ioooC. and Allen and Street 3 
had suggested, on the basis of an unpub 
lished research by T. Hewitt, not over 
0.005 P er cent oxygen at 5ooC. 

In the introductory stages of the inves 
tigation indirect evidence accumulated to 
suggest that the solubility limits were 
significantly lower than those previously 
reported. For this reason it was deemed 
essential to clarify this point before pro 
ceeding with a program intended to estab 
lish a quantitative relationship between 
hydrogen embrittlement and oxygen con 
centrations positively known to be within 
the solid solution range. Furthermore, 
the increasing production and the highly 
selective applications of the commercial 

Prom a part of a dissertation presented by E. N. 
Skinner, Jr, to the Faculty of the School of Engineer 
ing, Yale University, in partial fulfillment of the 
requirements for the degree of Doctor of Engineering. 
Manuscript received at the office of the Institute 
Dec. 2, 1940. Issued as T.P. 1280 in METALS TECH 
NOLOGY, January 1941. 

* Professor of Metallurgy, Yale University, New 
Haven, Conn. 

t Metallurgist, Research Laboratory, The Inter 
national Nickel Co., Inc., Bayonne, N. J. 

1 References are at the end of the paper. 



" oxygen-free" coppers inevitably direct 
attention to the practical necessity for exact 
knowledge regarding the effects of minute 
concentrations of the elements ordinarily 
found in the high-purity coppers now 
available. Of these elements oxygen oc 
cupies a unique position, not only by virtue 
of its specific effects on copper but more 
particularly in view of its important reac 
tive tendencies toward the other impurities 
with which it is associated. 

Rhines and Mathewson considered sev 
eral methods for measuring the solubility 
of oxygen and finally adopted the simple 
plan of saturating the specimens, obtained 
from redeposited cathode copper, by heat 
ing at the desired temperature in air. After 
the removal of surface oxide by mechanical 
and chemical means, the oxygen content 
was determined in the underlying metal. 
In general, this process of saturation, with a 
few modifications, was followed by the 
present writers. Although the procedure is 
direct and has obvious advantages, its 
effectiveness, unfortunately, is reduced by 
an excessive scaling of the copper, particu 
larly at the higher temperatures. After 
saturation treatments above 9ooC. the 
writers found that the specimens, neces 
sarily rather thin in the original condition, 
consisted largely of scale and yielded too 
little metal for a satisfactory analytical 
sample. Furthermore, great difficulty was 
experienced in effecting complete removal 
of the scale. Needless to say, it is necessary 
to exercise extraordinary care in the prepa 
ration of the surface of the analytical 
sample in order to eliminate submicro- 
scopic fragments of cuprous oxide which 
ordinarily are lodged in the crevasses and 



301 



302 



SOLUBILITY OF OXYGEN IN HIGH-PURITY COPPER 



surface irregularities of copper oxidized at 
high temperatures. Scaling of a serious 
nature was encountered by Rhines and 
Mathewson and they stated that a trouble 
some source of error was introduced by the 
surface conditions of their saturated 
material. 

EXPERIMENTAL PROCEDURE 

In order to reduce the loss of metal by 
oxidation during the high-temperature 
treatments, the time of heating in air was 
short, simply that necessary to form a thin 
coat of oxide. After surface oxidation was 
effected the air was displaced by purified 
nitrogen, which served as a protective 
atmosphere during the remainder of the 
saturation period. By this procedure satis 
factory specimens were obtained and the 
loss of copper, including the loss incurred 
by the mechanical and chemical cleaning 
of the surface, did not exceed 10 per cent. 
Later experiments, in which copper was 
saturated at temperatures ranging from 
550 to 9ooC. in air, stagnate and circu 
lating nitrogen furnished excellent analyti 
cal checks. The oxygen content, determined 
by the indirect method of Bassett and 
Bedworth, 4 was considered to be the per 
centage loss-in-weight of the sample after 
prolonged heating in pure hydrogen. 
Sulphur was, of course, determined simul 
taneously and the necessary correction 
applied to the total weight loss. For the 
supply of hydrogen for the combustion 
train, purified tank hydrogen* was sub 
stituted for the conventional Kip generator, 
in order to eliminate the presence of 
acetylene apparently formed from carbon 
in mossy zinc. Otherwise the analytical 
procedure was essentially that followed by 
Rhines and Mathewson. 



* The gas was passed successively through 
hot copper chips, concentrated HsS04 satu 
rated with KaCrsO?, 10 per cent KOH satu 
rated with KMn0 4 , concentrated H S 4 0, stick 
KOH and finally dried by passing through 
granular Cads and glass wool containing PsOe. 



The copper used in the preliminary work 
was supplied by the United States Metals 
Refining Co. and consisted of O.F.H.C. 
copper of normal purity. From a 5o-lb. 
slab, bars i in. thick were obtained by 
planing off the upper and lower portions of 
the cake. These bars were cold-rolled* to 
o.io in., a reduction of 90 per cent, with no 
intermediate anneals. Pieces cut from the 
strips were then saturated with oxygen, as 
previously described, at a series of tem 
peratures ranging from 500 to io4oC. 
The surface scale was broken and largely 
removed by tapping with a light hammer. 
After the surface of the pieces had been 
cleaned with emery cloth, the metal was 
pickled in 1:1 nitric acid. The specimens 
were then plunged into a beaker of cold 
water which had been previously boiled to 
remove oxygen and carbon dioxide. After 
this rinse the copper was dried rapidly by 
rubbing vigorously with a nap-free absorb 
ent paper. Every effort was made to avoid 
superficial oxidation, as the surface layer 
was to be included in the chips taken for 
chemical analysis. The procedure described 
was found preferable to rinsing in ether 
and alcohol and eliminated the possibility 
of contamination by a residue from the 
organic solutions. 

All samples for analysis were in the form 
of chips removed by a slowly operating 
shaper. After screening to remove the fines, 
the chips were weighed and introduced into 
the ignition bulb with the minimum of 
delay, in order to reduce the error due to 
surface oxidation. The heat-treatments 
involved, together with the oxygen and 
sulphur determinations, are summarized in 
Table i and the solubility curve plotted 
from the data is shown in Fig. i, which 
also includes a reproduction of Rhines and 
Mathewson s curve. It is interesting to note 
that the sulphur content of the preoxidized 
specimens heated to 1000 and io40C. in 



* The authors are indebted to the Bridge 
port Brass Co. for the rolling of this material. 



ARTHUR PHILLIPS AND E. N. SKINNER, JR. 



303 



nitrogen was reduced from approximately 
0.003 per cent to about 0.0004 P e r cent. 

Since Rhines and Mathewson had used 
redeposited high-purity copper cathodes, 



on chemical and spectrographic methods, 
which are hardly effective in yielding 
satisfactory analytical results on coppers 
containing impurities in minute concentra- 




it was deemed imperative to extend the 
present investigations to coppers of various 
origins and impurity contents. Although 
this precaution is unquestionably desirable, 
an intelligent execution of the program was 
made difficult by the necessity of depending 



tions. However, in spite of admittedly 
questionable quantitative expressions of 
impurity content, it is believed that the 
following results have been obtained on 
coppers differing sufficiently in this respect 
to afford oxygen-solubility data applicable 



34 



SOLUBILITY OF OXYGEN IN HIGH-PURITY COPPER 



to the high-purity coppers of present day 
interest. 

A prior treatment of Raritan cathode 
copper designed to eliminate sulphur and 
certain other impurities by high-tempera- 



in Table i and the points are plotted in 
Fig. i. 

The writers were fortunate in securing 
from the American Smelting and Refining 
Co. a quantity of extremely high-purity 



TABLE i. Data on Solubility Limits of Oxygen in Copper 



Kind of Copper 


Tempera 
ture of 
Saturation 
Deg. C. 


Time of 
Saturation 
Hr. 


Atmos 
phere 
during 
Saturatio 


Oxygen, 
Per Cent 
by Weight 


Sulphur, 
Per Cent 
by Weight 


O.F.H.C. (as reed.) 














550 


203 


Air 


0.0018 


o . 0027 
o. 0026 




550 


1600 


Air 


0. 0017 


0. 0021 




700 


75 


Air 


0.0020 


0.0029 




700 
850 
850 


90 
35 
50 


Air 
Air 
Air 


O.OO2I 
O.O022 
0.0023 


0.0027 
O.OO32 
0.0030 




850 


65 


Air 


0.0025 


0.0030 




900 


21 


Nitrogen 


O.OO27 


o . 0030 


O.F.H.C / 


900 
900 


30 


Nitrogen 


0.0035 


0.0035 


\ 


950 


33 

40 


Nitrogen 


o . 0030 
0.0034 


0.0027 
0.0019 




1000 


T-5 


Air 


0.0054 


0.0033 




1000 


20 


Air 


0.0050 


o . 0033 




1000 


24 


Air 


0.0053 


0.0031 




IOOO 


20 


Nitrogen 


0.0055 


0.0029 




1000 
IOOO 


60 
65 


Nitrogen 
Nitrogen 


0.0049 
o . 0046 


0.0004 
0.0003 




1040 


35 


Nitrogen 


0.0070 


0.0004 




1040 


35 


Nitrogen 


0.0068 


. 0006 


Raritan cathode vacuum melted -J 


800 
950 


85 
65 


Nitrogen 
Nitrogen 


0.0042 
0.0046 


O.O002 
0.0002 




( 


1050 


30 


Nitrogen 


0.0077 




{" 


650 


210 


Air 


0.0014 






850 


55 


Nitrogen 


0.0029 






IOOO 


25 


Nitrogen 


0.0045 






IOOO 


2 5 
25 


Nitrogen 
Nitrogen 


o. 0053 
0.0055 






1050 


30 


Nitrogen 


0.0075 






1050 


30 


Nitrogen 


0.0072 




A.S. and R. containing 0.0219 per cent silver. .) 


700 

850 

IOOO 
IOOO 


168 
55 
30 
30 


Air 
Nitrogen 
Nitrogen 
Nitrogen 


0.0022 
0.0026 
O.OO43 

o . 0044 




( 


700 


145 


Air 


O.002I 




Rhines and Mathewson remelted micro- J 


700 


145 


Air 


0.0021 




specimens. ( 


850 


55 


Nitrogen 


O.0028 




1 


IOOO 
IOOO 


50 
50 


Nitrogen 
Nitrogen 


0.0044 
0.0045 





ture vacuum melting in association with 
pure graphite furnished a supply of copper 
containing not over 0.0002 per cent sul 
phur and 0.0008 per cent oxygen. Inci 
dentally, it is believed that the material 
used by Rhines and Mathewson for further 
refining came from this lot of copper. From 
the ingot so obtained, a strip o.io in. 
thick was prepared and pieces from it were 
saturated at 800, 950 and io5oC. The 
heat-treating schedule and the resulting 
oxygen determinations are also assembled 



copper in the form of %-in. rod. Purifica 
tion of the metal had been effected by two 
depositions from a sulphate electrolyte 
and a third from a nitrate solution. The 
resulting cathode material was melted in a 
pure graphite crucible and final solidifica 
tion in rod form was accomplished by extru 
sion of the melt through a water-cooled 
die surrounded by a reducing atmosphere. 
The analysis of this material, by a combina 
tion of chemical and spectrographic 
methods, was reported by the American 



ARTHUR PHILLIPS AND E. N. SKINNER, JR. 



Smelting and Refining Co. as follows: Se 
not detectable; Te not detectable; Si, nil; 
Al, nil; Mg, nil; Fe, Ni, Sb, As, Pb, Ag, 
Bi, Sn, if present less than o.oooi. S, o.oooi 
(max.) and O, 0.0004 (max.)* were deter 
mined by the authors. 

The copper rods were surface-oxidized, 
according to previously described practice, 
and all specimens, except in the 65oC. 
treatment, were saturated in an atmosphere 
of nitrogen. The data of Table i and the 
plotted points of Fig. i indicate that any 
differences in oxygen solubility between 
this pure copper and the less pure O.F.H.C. 
material is negligible in comparison with 
the experimental error involved. 

An opportunity was offered to make 
oxygen-solubility measurements on this 
high-purity copper to which 0.0219 per 
cent silver had been added to the melt by 
the American Smelting and Refining Co. 
That silver has little or no effect on the 
oxygen solubility is evident from the 
results assembled in Table i and Fig. i. 

An effort was made to locate some of the 
copper used by Rhines and Mathewson and 
we were successful in finding a number of 
specimens that had been used for their 
microscopic examinations. The copper had 
been in storage for almost seven years 
and naturally showed evidences of surface 
oxidation and sulphidation. After cleaning 
in a nitric acid solution the specimens were 
melted in a vacuum furnace to produce a 
slug weighing about 400 grams. This 
material was converted into chips, dis 
carding the outer skin. In lots of about 
100 grams the chips were charged into a 
combustion bulb and heated overnight at 
95oC. in contact with a stream of hydro 
gen. The several residues were pressed into 
a slug, which then was melted in a vacuum 
furnace in a previously "burned-in" 
alundum thimble. An analysis of a sample 
taken from the ingot showed no measurable 

* According to W. A. Baker of the British Non- 
ferrous Research Association, the superficial oxidation 
of low-oxygen copper introduces an error of 0.0002 
per cent. This correction, however, has not been 
applied to any of our oxygen determinations. 



concentration of sulphur and about 0.0006 
per cent oxygen. It may be assumed that 
the material was then comparable from 
the standpoint of purity to its condition 
when exposed to saturation treatments by 
Rhines and Mathewson. From the cast 
slug, strips* (o.io in. thick) were prepared 
and specimens subjected to saturation 
treatments. Our results on this material 
(Table i, Fig. i) are in good agreement with 
the data obtained on the other coppers. 

DISCUSSION or RESULTS 

It is difficult to account for the signifi 
cant differences in oxygen-solubility values 
made evident by a comparison of our re 
sults with the values given by Rhines and 
Mathewson. Experimental work of this 
nature requires constant vigilance in order 
to avoid contamination of the material, and 
and the final results are, unfortunately, 
based on analytical procedures that lack 
the necessary accuracy for the concentra 
tions under test. Until more adequate 
determinative methods are available, all 
data relating to minute amounts of im 
purities in copper must be regarded with 
suspicion and at best accepted only 
tentatively. This viewpoint is most cer 
tainly adopted by the present writers, who 
will be the last to claim that the solubility 
limits indicated in this paper are final and 
irrevocable. 

During the course of the investigation, 
an effort was made to obtain checks on our 
oxygen measurements by skilled operators 
in other laboratories. Unfortunately, the 
sensitivity demanded is distinctly beyond 
that required for routine analysis in the 
control laboratories of the copper industry 
and the research laboratories have not 
included this determination in their special 
analytical procedures. After the completion 
of our work, however, W. A. Baker, of the 

* The writers are indebted to the American 
Smelting and Refining Company for the 
following analysis of the strip material: Fe 
o 0008, Sb o.oooi, Pb 0.0003, Sn 0.00008, Ni 
0*0002, Bi Nil, As Nil, Si Nil, Ag o.i oz./T. 



SOLUBILITY OF OXYGEN IN HIGH-PURITY COPPER 



British Non-ferrous Research Association, 
kindly consented to make oxygen deter 
minations on some of our specimens. The 
method used by Baker has been described 
by him 5 and consists essentially of heating 
the massive copper specimen in pure 
hydrogen under reduced pressure, condens 
ing out the water vapor at 8oC., vapor 
izing the water in an evacuated system and 
determining the amount of oxygen by a 
measurement of the vapor pressure. 
Baker s values, together with the authors 
determinations on duplicate samples are 
shown in Table 2. Although the two sets of 
quantitative results, based on distinctly 
different analytical procedures, are not in 
absolute agreement, they furnish supple 
mentary evidence regarding the magnitude 
of oxygen solubility previously indicated 
by this work. 

TABLE 2. Determinations of Oxygen by 

Two Investigators 

PER CENT 



Specimen, O.P.H.C. 


Author s 
Deter 
mination 


Baker s 
Deter 
mination 


Saturated 1600 hr. in air at 550C. 
Saturated 73 hr. in air at 7OOC. . . 
Saturated 35 hr. in nitrogen at 
iO40C 


0.0017 

O.OO20 
O.OO68 


0.0008 

O.OOIO 

o. 0053 









Assuming that Baker s results constitute 
a reasonably satisfactory confirmation of 
the oxygen- con tent measurements, it be 
comes necessary to look elsewhere for the 
explanation of the discrepancy between our 
results and those of earlier workers. It is, 
of course, obvious that incomplete satura 
tion of the specimens could account for the 
lower values reported here. It may be 
stated in this regard that our shortest 
heating periods, as suggested by prelimi 
nary observations, were longer than we 
deemed necessary and the longer times 
adopted as a desirable precaution gave 
assurance that equilibrium had been 
attained. Furthermore, the saturation 
treatments satisfied the time requirements 



as calculated from Ransley s 6 diffusion 
data. 

It is conceivable that the physical con 
dition of the specimens used by Rhines and 
Mathewson may account for their higher 
values. Reference has already been made to 
the difficulty which they encountered in 
ridding the metal of the heavy scale, par 
ticularly after the treatments at high tem 
peratures. The difficulty in accepting this 
observation as a satisfactory explanation 
of the discrepancy is obvious, however, 
when it is noted that in both investigations 
air served as the saturation atmosphere for 
temperatures up to about 9ooC., resulting, 
presumably, in similar scale formations. At 
the higher temperatures the nitrogen 
atmosphere undoubtedly furnished more 
satisfactory material for analysis and the 
differences in solubility for these tempera 
tures could be explained adequately by the 
improved procedure for saturation. Still 
another consideration may be based on the 
surface condition that results from the 
rolling of a small ingot. Mathewson and 
Rhines used relatively small castings and 
it is possible that the rolled strip had an 
open texture due to a limited reduction of a 
cast material. The saturation treatment of 
such a strip would be accompanied by 
oxidation of minute fissures from which 
the oxide particles could not be removed by 
ordinary surface-cleaning methods. 

Although it is common practice to sub 
ject experimentally derived constitutional 
data to further verification by supplemen 
tary experimental procedures, some atten 
tion has been devoted to the testing of 
such data on the basis of their conformity 
to thermodynamic considerations. Fink and 
Freche 7 were conspicuously successful in 
showing that the experimentally deter 
mined solid solubility curves of several 
binary aluminum systems conformed to the 
laws of perfect solutions in that log* x = 

TT 

~~Tlf ~^~ ^" ^ ccor( ^ n S to this equation, if 
the log of the atomic per cent of the solute 



ARTHUR PHILLIPS AND E. N. SKINNER, JR. 



307 



is plotted against the reciprocal of the 

absolute temperature a linear relationship 

must result if the solution behaves ideally. 

In studying gas-metal equilibria, Borelius 



intersecting lines better emphasize, and 
perhaps more clearly express, the deviation 
from linearity at temperatures above 
925C. It is also interesting to note that a 



/uu 
.090 
.060 

syjfi 








































A 












050 
040 

030 

| m 

\ 

^ 0/0 
$ 009 
K 006 
^ 007 

& 6 
| 5 
004 

003 
Q02 

001 


\ 












\ 












\ 


\ 


^-^ 


-^^_ 






A 








A ^~~ 


^ 


\ 


"w 












\ 


^~ - .. 














^-^-^^^_ 












* 


^_ 














*~~ 




































A 






G/=:y 


VC Copper 
7#s a/7</ Mcrfhewson 
er s Ana/ytica/ ffesu/fs 

700 600 

\ \ 




/ooo 4 
\ 


C 900 

\ 


A Bak 

600 

\ 


soo 

\ 


r 8 9 /O // /2 S3 



FIG. 2. SOLID SOLUBILITY OF OXYGEN IN COPPER. 
Log of atomic percentage of oxygen plotted against 



and Lindholm 8 plotted the results of 
Sieverts and his co-workers on a number of 
hydrogen-metal systems and, in each case, 
complete conformity to the linear relation 
ship was observed. There is little evidence 
available in this regard on oxygen-metal 
systems, only for silver 9 and cobalt 10 has 
this relationship been emphasized. 

This criterion has been applied to the 
solubility data of Table i with the results 
shown in Fig. 2. It is evident that the points 
do not lie on a straight line. Although it is 
possible to draw a reasonably smooth curve 
through the points, it is believed that two 



plot of Rhines and Mathewson s results 
suggests a comparable deviation although 
their data furnished too few high-tempera 
ture values to be considered as other than 
possibly supplementary evidence. Again 
the analytical results reported by Baker 
offering only three points, to be sure fail 
to fall on a straight line. It is, of course, 
futile to discuss at this time the significance 
of these plots, since they are based on the 
questionable assumption that we are 
dealing with an ideal solution. Neverthe 
less, it may be suggested, but not stressed, 
that the point of intersection may indicate 



3 o8 



SOLUBILITY OF OXYGEN IN HIGH-PURITY COPPER 



a difference in type of solution above and 
below a critical temperature. 

Although the present work is concerned 
primarily with the solubility relationships 
between copper and oxygen, it is realized 
that the several coppers used, relatively 
pure as they were, contained many im 
purities in varying concentrations. All of 
the metallic elements found in copper are 
capable of forming oxides and, with the 
exceptions of silver and any precious metals 
present, the oxides have higher free energies 
of formation than cuprous oxide and are, 
therefore, more stable than that oxide. 
Under the conditions for ensuring complete 
saturation of the copper, equally favorable 
conditions are attained for the internal 
oxidation of metal impurities. Rhines 11 has 
reported evidence of internal oxidation in 
about 35 copper alpha-solution alloys. 
When we realize, however, that the total 
metallic impurities of the special lot of 
American Smelting and Refining copper 
total less than o.ooi per cent, it is evident 
that even on the assumption that all of 
the potential oxides are absolutely insoluble 
in oxygen-saturated solid copper, con 
siderably less than o.ooi per cent oxygen 
would be segregated as oxide and, there 
fore, not properly included in the oxygen 
content required for saturation. Obviously, 
any correction for the oxides present in 
high-purity copper may be considered 
negligible until more refined analytical 
methods are available. 



1. F, 

2. D 



4. W 

5- W 

6. C. 

7. W 



REFERENCES 

, X. Rhines and C. H. Mathewson: Solubility of 
Oxygen in Solid Copper. Trans. A.I.M E. 
(1934) "I 337- 

>. Hansen, C. Marryat and G. W. Ford: Inves 
tigation of the Effects of Impurities on Copper 
IThe Effect of Oxygen on Copper. Jnl. Inst. 
Metals (19.23) 30, 197. 

. P. Allen and A. C. Street: An Investigation of 
the Effects of Hydrogen and Oxygen on the 
Soundness of Copper-Nickel Alloys. Jnl. Inst 
Metals (1933) 51, 233. 

. H. Basse tt and H. A. Bed worth: Estimation of 
9 x /f ?" and Sul P h " r in Refined Copper Trans 
A.I.M. E. (1926) 73, 784. 

vA , Ba i< e r: The Estimation of Oxygen in 
Metals. Jnl. Inst. Metals (1939) 65, 34? 

E. Ransley: The Diffusion of Oxygen in 
Copper. JnL Inst. Metals (1939) 65, 147. 
. L. Pink and H. R. Freche: Correlation of 
Equilibrium Relations in Binary Aluminum 



Alloys of High Purity. Trans. A.I.M. E. (1034) 
111,304. 

8. Borelms and Lindholm: Diffusion von Wasser 

stoff durch Metallen. Ann.Phys. (1927) 82,201. 

9. Steacie and Johnson: The Solubility and Rate of 

Solution of Oxygen in Silver. Proc. Roy. Soc. 
(London) (1926) 112, 542. 

10. A. U. Seybolt and C. H. Mathewson: The 
Solubility of Oxygen in Solid Cobalt. Trans 
A.I.M.E. (i93S) H7, 156. 

rr. F. N. Rhines: A Metallographic Study of Inter 
nal Oxidation in the Alpha Solid Solutions of 
Copper. Trans. A.I.M.E. (1940) 137, 246. 

DISCUSSION 
(E. M. Wise presiding] 

J. S. SMART, JR.,* Barber, N. J. I have 
had the pleasure of following the progress 
of the present work in the experimental stages, 
and am convinced that in this very condensed 
paper the authors have presented the results 
of an extremely laborious and skillfully con 
ducted research. The excellent agreement 
obtained from coppers of diverse composition 
seems certain to negate the possibility of a 
relation between the oxygen solubility and the 
small amounts of impurities normal to com 
mercial coppers. 

The experimental method employed by the 
authors is one of long standing and proven 
worth for oxygen contents in the tough-pitch 
range. With the minute quantities involved in 
the present problem, however, meticulous care 
is essential to obtain even consistent results. 
In this, the authors have been conspicuously 
successful, and in all probability their results 
represent the present method at its best. 

Unfortunately, the procedure involves the 
conversion of the sample to small chips having 
an enormous surface area, and there is no ques 
tion that a measurable surface pickup is ob 
tained. Heuer 12 found the effect to approximate 
o.ooi per cent O 2 , and this would appear to 
rationalize the differences of o.ooi to 0.0015 per 
cent between the results of the authors and 
those of Baker (who uses the sample in the 
form of strip), were it not for the fact that 
Baker checked one of his determinations with 
chips and found the additional surface effect 
to be only 0.0002 per cent O 2 . Consequently, 
it appears that the discrepancies may be partly 
dependent on a now obscure difference between 
the two methods, in addition to the superficial 
surface oxidation effect. 



* Central Research Laboratory, American Smelting 
and Refining Co. 

12 R. P. Heuer: JnL Amer. Chem. Soc. (1927) 49, 



DISCUSSION 



Among the sources of difficulty, the purity 
of the reagent hydrogen appears due for more 
careful consideration than it has been given 
heretofore. The insertion of a palladium 
thimble in the hydrogen stream, and the use 
of a liquid-air trap, would certainly result in a 
considerable improvement over the usual 
purification train, while preliminary ignitions 
in carbon monoxide may provide access to the 
elimination of variable surface pickup. 

From the evidence now at hand, it seems 
probable that the present oxygen solubility 
results, while much lower than those previously 
determined, are still slightly on the high side, 
probably by an amount that is nearly constant 
at all temperatures. It would also appear that 
the maximum deviation is about o.ooi to 
0.0015 per cent 0%. The authors results may be 
subject to such future correction, as they 
themselves suggest, but it is certain that the 
limitations of the present method must first 
be overcome, or a new procedure employed, if 
significant improvement is to be effected. 

W. B. PRICE,* Waterbury, Conn. On Fig. i, 
the Raritan cathode shows consistently a 
larger volume of oxygen than the other types 
of copper, except at the end of the curve. This 
is particularly noticeable at 800, where there 
is twice the amount of oxygen. Was the sulphur 
removed from the cathode, and if it was not 
might not the sulphur account for that increase 
in oxygen? 

E. N. SKINNER, JR. Mr. Smart s comments 
should be considered a very necessary corollary 
to this paper. As we have stated before, we 
feel that this method of analysis is severely 
burdened at these low quantities of oxygen. 
A method similar to that used by Baker which 
minimizes the adsorbed oxygen on the surface 
of the chemical sample would indeed be pref er- 
ble. That there was always some adsorption 
was fairly evident since we were never able 
to prepare a copper that analyzed zero per cent 
oxygen. Furthermore, the appearance of copper 
chips after heating in hydrogen, as compared 
with chips machined from a solid strip which 
had been given a long-time hydrogen anneal, 
can be detected instantly by the difference 
in color. 



As to the method of hydrogen purification, 
that, too, may be vastly improved. The use 
of a palladium thimble in conjunction with 
some system for freezing out residual impurities 
deserves consideration. 

F. N. RHINES,* Pittsburgh, Pa. An expres 
sion for the rate of internal oxidation of dilute 
copper alloys has recently been developed; this 
expression involves the solubility of oxygen in 
copper, which, if all the other variables are 
known, can be calculated by its use: 



B 



where 



The depth of oxidation is x in time t; D and 
D m are, respectively, the diffusion coefficients 
of oxygen and the alloying element in copper; 
Co is the solubility of oxygen in pure copper; C m 
is the concentration of the alloying element; 
and 0/M is the atomic weight ratio of oxygen 
and metal in the oxide formed. The rates of 
oxidation have been carefully measured in a 
large series of alloys and the other terms of the 
expression are well known, except C and D . 
Ransley has determined D in terms of C , 
so that the only value in doubt at the moment 
appears to be C . The approximate equality 
indicated in the second expression represents 
precision to the first two significant figures. 

By substituting in this expression Phillips 
and Skinner s value and again Rhines and 

X 2 

Mathewson s value and solving for which 

t 

has been experimentally determined, a test 
of the validity of the two C values may be 
obtained : 




.X" 2 



-y calculated from Phillips and Skinner s data 

- 8.28 X ro~ 7 

calculated from Rhines and Mathewson s 

t 

data = 6.66 X icr 7 

experimentally determined = 6.56 X io~ 7 

These values refer to a 0.103 per cent Si alloy 
oxidized at ioooC., and are typical of a series 



* Chief Chemist and Metallurgist, Scovill Manu 
facturing Co. 



* Assistant Professor of Metallurgy, Carnegie 
Institute of Technology. 



3 io 



SOLUBILITY OP OXYGEN IN HIGH-PURITY COPPER 



of cases tested. If this method is valid (the 
indirect approach does leave some room for 
doubt), it indicates that the solubility of 
oxygen in copper is slightly higher than that 
reported by Rhines and Mathewson. 

W. A. BAKER,* London, England, In the 
footnote on page 305, I am loosely quoted as 
stating that "the superficial oxidation of low- 
oxygen copper introduces an error of 0.0002 per 
cent." Actually, I pointed out that my analyses 
on massive samples gave figures lower than 
the authors results of analyses on chips, the 
difference being constant (within the probable 
limits of experimental error) at o.ooi per cent 
oxygen. I suggested that this was due to 
superficial oxidation, which would naturally be 
greater on the chip samples, and indicated that 
we had observed that superficial oxidation of a 
massive specimen of lead might introduce an 
error of 0.0002 per cent. There is a further 
source of error, which is also aggravated by the 
use of samples with a large surface area -weight 
ratio; namely, the adsorption of water vapor 
on the surface of the specimen prior to its 
insertion into the combustion tube. If the 
authors results are in fact consistently o.oor 
per cent oxygen high from this source of error, 
the location of the solubility curve is materially 
different from that shown, the error being 
equal to more than 100 per cent at the lowest 
temperatures considered. 

With regard to the disparity between the 
authors results and those of Rhines and 
Mathewson, it is conceivable that superficial 
oxidation and adsorption of moisture might 
account for the observed difference. It is worthy 
of note that whereas the authors weighed the 
chips and then introduced them into the 
ignition bulb with a minimum of deky, Rhines 
and Mathewson weighed the chips in the 
ignition bulb and, owing to the bulk of the 
latter, found it necessary to allow the sample 
and bulb to stand for an hour prior to final 
weighing. Any increase in weight due to 
adsorption of water prior to reduction was 
probably not compensated for by adsorption 
after reduction, since in the subsequent cooling 
period the chips were probably held in hydro 
gen, to avoid surface oxidation. The magnitude 



* * Research Department, British Non-ferrous 
Metals Research Association. 



of the error that might result from adsorbed 
moisture can be approximately estimated thus. 
We have determined the amounts of water 
vapor adsorbed on the surface of aluminum 
specimens and have found amounts of the 
order of 0.5 to i mg.*per 100 sq. cm. Assuming 
that copper adsorbs similar amounts of water 
and that the chips used were, say, rectangular 
blocks, i by i by 2 mm., the surface area of 
50 grams of such material would be of the 
order of 200 sq. cm. and the consequent error in 
the oxygen estimation by difference in weight 
could then amount to 0.004 per cent oxygen. 

The writer is by no means sure that such 
errors account for the discrepancies noted, but 
it does appear that these sources of error should 
be carefully considered when making analyses 
for minute traces of oxygen. 

A. PHILLIPS AND E. N. SKINNER, JR. (authors 
reply). With respect to the method of satura 
tion, we heated the specimens for periods that 
gave constant end points and considered the 
diffusion rates involved in connection with the 
evidence made available by the studies of 
Rhines and Mathewson and Ransley. We are 
not particularly concerned with the exact 
nature of the oxidized surface of the specimens; 
i.e., whether it consists essentially of CuO or 
CiizO, or both. The oxide surface simply served 
as a source of oxygen and there is abundant 
evidence proving that oxygen diffused to a 
constant concentration after heat-treatments 
extending beyond a given minimum time. 

We are in agreement with Mr. Smart s 
opinion that the present results are probably 
a little on the high side. W T e were particularly 
concerned with the error involved in the use 
of small chips for analysis and concluded that 
the error introduced could not be stated in 
absolute terms, but was highly dependent on 
the many conditions associated with the 
machining and handling of the chips. 

Answering Mr. Price s query regarding the 
higher concentrations shown for the Raritan 
copper, we do not believe that it is in any way 
due to the sulphur content, which was low. 
We have no explanation for the apparent 
difference in oxygen solubility between this 
copper and the other kinds and do not wish 
to place any emphasis on the implied difference. 
Dr. Rhines has attempted an evaluation of 
our results and those previously published by 



DISCUSSION 



K bines and Mathewson based on a mathe 
matical treatment of data furnished by studies 
of internal oxidation rates. We are unable to 
state at this time whether the agreement 
derived by inserting the Rhines and Mathew 
son data in the formula is conclusive or is, on 
the other hand, merely a coincidence. 

We are somewhat puzzled regarding Mr. 
Baker s comments on the error due to super 
ficial oxidation. In his original letter giving the 
results of his determinations on our specimens, 
Mr. Baker made the following statement: 
"Our analyses above were made on small strips 



cut from the samples supplied, the sample being 
first freed from superficial oxide films with a 
clean file. Sample Ws was also analyzed using 
chippings obtained by milling the previously 
cleaned strip. The chipping yielded 0.00099 per 
cent oxygen" (as against 0.00079 per cent for 
the cleaned strip). "The difference (0.0002 per 
cent) between the oxygen contents of the 
strips and chip samples from W$ may reason 
ably be ascribed to superficial oxidation of the 
surfaces of the chips and it appears probable 
that your analytical results are consistently 
high to this slight extent." 



Hydrogen Embrittlement of Pure Copper and of Dilute Copper 
Alloys by Alternate Oxidation and Reduction 

BY FREDERICK N. RHINES,* MEMBER A.I.M.E. AND WILLIAM A. ANDERSON! 

(Cleveland Meeting, October 1940) 

THE investigations of Wyman 1 have and titanium may themselves be reduced 
imonstrated that copper deoxidized with by hydrogen when in contact with copper: 



demonstrated that copper deoxidized with 
several of the commonly used agents that 
confer immunity to ordinary hydrogen em- 
brittlement can still be embrittled if it is 
annealed alternately under oxidizing and 
reducing conditions. During the oxidizing 
cycle, oxygen diffuses into the copper and 
frequently deposits a subscale composed of 
the oxide of the residual deoxidizing agent. 
Upon subsequent reduction with hot hydro 
gen, the copper becomes ruptured along 
the grain boundaries to a depth sharply 
limited by the previous penetration of oxy 
gen as delineated by the inner boundary of 
the subscale. 

Obviously the sensitivity to this kind of 
hydrogen embrittlement is engendered by 
the introduction of oxygen during the oxi 
dizing anneal, but beyond this the mecha 
nism of the process, and hence the 
controlling factors, remain obscure. There 
are several possibilities: (i) oxygen dissolved 
in the copper matrix of the subscale may, 
upon cooling prior to the hydrogen anneal, 
precipitate as cuprous oxide which is subse 
quently reduced by hydrogen as in the 
familiar case of hydrogen embrittlement; 
(2) oxygen in solid solution in copper may 
itself confer susceptibility to embrittlement 
by hydrogen; (3) the normally refractory 
oxides such as those of aluminum, silicon 



Manuscript received at the office of the Institute 
June 17, 1940. Issued as T.P. 1235 in METALS TECH 
NOLOGY, October 1940. 

* Assistant Professor of Metallurgy, Member 
Metals Research Laboratory, Carnegie Institute of 
Technology, Pittsburgh, Pa. 

t Research Assistant, Metals Research Laboratory, 
Carnegie Institute of Technology, Pittsburgh, Pa. 

1 References are at the end of the paper. 



by hydrogen when in contact with copper; 
or (4) the precipitated oxides of the sub- 
scale may be complex compounds such, 
perhaps, as Si02.Cu2O, which are partially 
reduced by hydrogen. These possibilities 
have been examined, and it has been con 
cluded that the first three and possibly the 
fourth are operative collectively or indi 
vidually under more or less predictable 
conditions. 

EFFECT or PRECIPITATED CUPROUS OXIDE 
AND DISSOLVED OXYGEN 

The literature provides no definite proof 
that oxygen in solid solution in copper can 
confer susceptibility to hydrogen em 
brittlement ; indeed, the research of Leiter 2 
seemed to indicate the opposite that to 
produce the effect, cuprous oxide must be 
present in massive form and in a more or 
less continuous network along the grain 
boundaries. To test this point, pure copper 
has been annealed in hydrogen under con 
ditions designed to ensure the presence in 
various samples of : (i) precipitated cuprous 
oxide, (2) dissolved oxygen only, and (3) 
no oxygen in any form. Very pure copper* 
said to contain no oxygen and no spectro- 
scopically detectable traces of any other 
element (that is, no individual detectable 
impurity in quantity greater than o.oooi 
per cent) was used. Strips of this material 
cold-rolled 75 per cent to J in. in thickness 
were subjected to the following treatments, 
the results of which are noted: 



* Obtained through the courtesy of Dr. A. J. 
Phillips, of the American Smelting and Refining Co. 



312 



FREDERICK N. RHINES AND WILLIAM A. ANDERSON 



313 



1. Embedded in a mixture of equal parts 
of cuprous oxide and copper powder, a bar 
was heated to 8ooC. for i6H hr. in an 
atmosphere of stagnant nitrogen, cooled in 
the furnace, and then reheated to 800 C. 
for 2 hr. in dry hydrogen. This specimen 
withstood only five right-angle bends before 
breaking, and upon microscopic exami 
nation was found to have a considerable 
number of moderately large holes along the 
grain boundaries (Fig. 16). A section of this 
sample taken after oxidation and before 
the hydrogen treatment was found to con 
tain a precipitate that exhibited the typical 
cuprous oxide color in its larger particles 
(Fig. i a). In a similar experiment carried 
out at 1000 C. similar results were ob 
tained; the specimen withstood less than 
one complete right-angle bend, and the 
microstructure showed the presence of 
numerous large cavities at the grain 
boundaries. Since hydrogen embrittlement 
begins at 400 C. and the sample was in 
contact with hydrogen while heating to 
8ooC., it is highly improbable that the 
precipitate was redissolved before em 
brittlement set in. This experiment is taken 
to indicate that the presence of precipitated 
cuprous oxide makes copper susceptible to 
hydrogen embrittlement. 

2. Embedded in the copper-cuprous 
oxide mixture, another sample was heated 
to 8ooC. for 45 hr. in a stagnant nitrogen 
atmosphere and then, without any lowering 
of the temperature, 2 hr. in dried hydrogen. 
This treatment should ensure the satura 
tion of the copper with oxygen, and, since 
the temperature was not allowed to de 
crease between treatments, there should be 
no precipitate of cuprous oxide. The sample 
withstood 2oK right-angle bends before 
breaking; this is somewhat less than the 
breaking count for sound annealed copper. 
There were a number of small holes at the 
grain boundaries (Fig. ic). Similar experi 
ments in which the oxygen was introduced 
by first producing a thin scale on the copper 
instead of using a bed of cuprous oxide 



powder gave similar results in treatments 
at both 800 and ioooC. In some instances 
the embrittlement was more severe, and a 
few specimens withstood less than two 
right-angle bends. Apparently oxygen in 
solid solution in copper produces a mild 
susceptibility to hydrogen embrittlement. 

3. Embedded in the copper-cuprous 
oxide mixture, another bar was heated to 
8ooC. for 45 hr. in stagnant nitrogen; then, 
without lowering of the temperature, 4 hr. 
in dry carbon monoxide; and, again without 
lowering of the temperature, 2 hr. in dry 
hydrogen. Ransley 3 has demonstrated that 
pure dry carbon monoxide does not pene 
trate copper appreciably and will withdraw 
the oxygen from copper at a rapid rate 
without causing embrittlement ; this sample 
should, therefore, be substantially free from 
oxygen before the hydrogen treatment. 
Various samples treated in this way with 
stood right-angle bends numbering as high 
as 33 before rupture. The microstructure 
was completely free from cavities of any 
kind (Fig. id)] there has been no em 
brittling action. 

4. A final sample was heated in dried 
hydrogen at 8ooC. for 2 hr. The structure 
was perfectly sound, and the resistance to 
the bend test was of the same order of 
magnitude as that of the samples treated 
by procedure 3 above; i.e., the copper was 
not embrittled. 

In summary, these observations indicate 
that a degree of susceptibility to hydrogen 
embrittlement is conferred upon copper 
either by oxygen in solid solution or by 
cuprous oxide precipitated from solid solu 
tion. The effect of the latter is evidently 
much the more severe. By heating in pure 
dry carbon monoxide, dissolved oxygen can 
be sufficiently withdrawn from saturated 
copper to destroy its sensitivity to hydro 
gen embrittlement. The copper used was 
not, in its initial state, embrittled by 
hydrogen. 

The equipment employed in these experi 
ments consisted of a fused silica tube fur- 



3 14 HYDROGEN EMBRITTLEMENT OF COPPER AND COPPER ALLOYS 



***.< 

i 







V 



9 
*> 



FIG. i.- PURE COPPER. 

Pure copper heated i6H to. at 8ooC. in a mixture of copper and cuprous oxide powder 
a closed nitrogen-filled chamber; furnace-cooled. Shows a precipitate of cuprous oxide. X 500. 
b Same as a reheated for 2 hr. at 8ooC. in an atmosphere of dry hydrogen. Shows mild hydro- 



c Pure copper heated 45 hr. at 8ooC. in a copper-cuprous oxide mixture in stagnant nitro 
gen, then without lowering the temperature, 2 hr. in dry hydrogen. Shows hydrogen embrittlement. 

5 <fSame as c but heated 4 hr. in pure dry carbon monoxide after oxidation and before introduc 
tion of the hydrogen. Shows no embrittlement. X S- 



FREDERICK N. RHINES AND WILLIAM A. ANDERSON 



3IS 



nace with a temperature controller. No 
metal other than the copper was introduced 
into the furnace; porcelain boats were used. 
Hydrogen was prepared by the action of 
hydrochloric acid upon zinc and was dried 
progressively with caustic soda, calcium 
chloride, and phosphorus pentoxide. Car 
bon monoxide was prepared by the dehy 
dration of formic acid with sulphuric acid 
and was freed from moisture and carbon 
dioxide with soda lime, phosphorus pen 
toxide, and a liquid-air trap. The tank 
nitrogen used for neutral atmospheres was 
deoxidized by passing it over heated copper 
foil. The gas flow was never fast enough to 
alter sensibly the temperature of the 
furnace. 

For the bend tests, a special vise was 
constructed. The lips of the jaws were 
rounded to a radius of Jfs i&* an d the 
protruding end of the sample was clamped 
to a pointer that moved over a protractor, 
so that the angle at which cracking or 
rupture occurred could be read within 5 of 
angle. The samples were bent first 90 in 
one direction then 180 in the opposite 
direction, and so on, each complete cycle 
being counted as four right-angle bends. 
Test pieces M in. thick and approximately 
y by 2 in. in area were used. This test, 
though crude, has been found by others to 
be very sensitive and, on the whole, a 
better indicator of the degree of em- 
brittlement than more elaborate tensile 
measurements. 

EFFECT OF FOREIGN OXIDES 

A series of dilute copper alloys, originally 
made for a study of internal oxidation, 4 was 
used. Cold-rolled strips of the original 
unoxidized alloys were cut into bars Jg by 
y by 2 in. and were subjected to three 
kinds of treatment: (i) embedded in a 
copper-cuprous oxide mixture, a set of 
alloys was heated 45 hr. at 8ooC. in a 
closed chamber; (2) another set was heated 
in dried hydrogen at 8ooC. for 2 hr. (the 
same treatment as paragraph 4 above); 



and (3) a third set embedded in a copper- 
cuprous oxide mixture was treated suc 
cessively at 8ooC.* with stagnant nitrogen 
for 45 hr., pure dry carbon monoxide for 
4 hr., and dried hydrogen for 2 hr., without 
at any time any lowering of the tempera 
ture (the same as treatment paragraph 3 
above). The first treatment was designed 
to produce a subscale of the alloying ele 
ment, the second to test the effect of 
hydrogen alone on the alloys in their initial 
state, and the third to produce a subscale 
and then to expose it to hydrogen after the 
removal from the copper of dissolved oxy 
gen that might mask the effect of the 
precipitated oxides by itself inducing hy 
drogen embrittlement. 

All bars were subjected to a bend test, 
the results of which are recorded in Table i. 
Since the embrittlement was invariably 
limited to the zone of internal oxidation 
that usually extended only part way 
through the samples, the appearance of the 
first crack was a better indicator of em 
brittlement than the final rupture; where 
fore, the angle of bend at which cracking 
was first observed has been recorded as 
well as the angle at rupture. When the 
sample was oxidized throughout, the first 
crack and rupture occurred almost simul 
taneously. The reproducibility of these re 
sults was remarkable despite the apparent 
crudity of the bend test and the variations 
to be expected in successive annealing 
treatments. Often the test results agreed 
within 5 of angle in successive experi 
ments. The over-all precision is felt to 
justify reporting the results within one-half 
of a right-angle bend in the general case 
and to one-quarter of a right-angle bend 
when brittleness was extreme. 

Table i shows that many of the alloys 
are embrittled to some extent by the 



* All of the experiments reported here were carried 
out at a temperature of 800 C. because previous 
studies have shown that hydrogen embrittlement is 
often, most severe at this temperature. Many tests 
run at temperatures of 900 and ioooC., not re 
ported, have given entirely similar results except that 
the action is more rapid at the higher temperatures. 



3 i6 



HYDROGEN EMBRITTLEMENT OF COPPER AND COPPER ALLOYS 



presence of the subscale itself. This is 
indicated chiefly by the early appearance 
of surface cracks. In some instances, 



seems to have imparted brittleness. The 
most pertinent observation, however, is 
that almost without exception the alloys 



TABLE i. Summary of Results 





Internal Oxidation 
Only 


Annealed in Hydrogen 
Only 


Internally Oxidized 
and Annealed in Ha 


_ Type of 


Alloy Composition, 
Per Cent, Balance 
Copper 


Number of 90 Bends 


Number of 90 Bends 


Number of 90 Bends 


Microstructure 
in Oxidized 
and Reduced 
Samples 




ist Crack 


Rupture 


ist Crack 


Rupture 


ist Crack 


Rupture 




0.03 Al 
0.17 Al 
0.03 As 
0.05 B 


6 

6*2 

20 


ii 


15 

12 
I 

19 


22 

16 

25 


2*1 
2 
1 


2H 
5*2 
2 
* 2 


III 
IV 
III 
II 


0.054 Be 
o.ioi Be 


5 
3 


IO 

6 


12 
4? 2 


14 
IS 


2*1 


9 

7 


IV 

IV 


0.03 Bi 


18 


20 


I 


1*4 


2 


2 ^ 


III 


Traces Ca 


IO 


12 3 


23 


*2 


4/2 


IV 


0.04 Cb 
Traces Ce 


12 

4 


20 
16 


12 

16 


17 

20 


I 


2 


III 


0.14 Co 


1 1 


23 


I 


I 


*2 


X2 


II 


o.io Cr 


IS 20 


*2 


*2 


2 


2 


II 


0.56 Fe 
0.03 Ga 


18 


21 


6 

12 


9 

12*2* 


I 


II 


IV 
V 


0.25 In 


7 


16 


17 


19 


*2 


*2 


I 


0.02 Li 


12 


14 


21 


28 


2 


3 


V 


o.i i Mg 


4 } 




IS 


24 


1*2 


2 


V 


0.033 Mn 




25 


12 


IS 


*2 


I 


I 


0.22 Mn 


3 


6 


IS 


21 


*2 


2 


I 


1.55 Mn 


2 


8 


16 


19 


*2 


8 


I 


0.115 Ni 
5 Ni 


II 
2*2 


14 

2/ 2* 


6 
16 


15 

17 


I 


I 


I 


0.03 P 


9 


IO 


ii 


16 


*2 


*2 


I 


0.24 P 




IO 


IO 


14 


I 


12 


I 


0.42 P 


1 3 


6? 2 


9 


13 


% 


7 


I 


0.03 Pb 


} 2 


I O l , 2 


20* 


26 


I 


15 


II 


0.03 Sb 


13 


16 


> 2* 


*^J 


*2 


*2 


III 


0.045 Si 
0.180 Si 




16*2 
2 


20 
3 


23 

14 


K 


8 /2 


II 

III 


0.858 Si 






4*2 


6 




24 


III 


0.31 Sn 


6 


II 


2 


2 


*2 


2 


I 


o.i Sr 


7 


13 


16 


21 


2 


*2 


III 


0.04 Ta 






5 


10 


*2 


I 


III 


0.05 Ti ! 8 


13 


5 




I 


I 


IV 


0.09 V 


9 


IO 


9 


17 


I 


2 


III 


0.21 Zn 


IS 16 


14 


17 


*2 


*2 


I 


0.16 Zr 


6 : 12 


12 


18 


5 


6 


V 


0.03 Be + 9-52 Zn 


1 2 


16 


21 


24 


ii 


20?- 2 


I 


0.13 Al + 9.29 Zn 


2 


12)2 


17 


20 


} 2 


12 


I 


0.085 Si -}- 9-8 1 Zn 


J 4 18 


16 


2O 


14 


18 


I 


0.006 Be + 4-93 Sn 
0.06 Al + 5-43 Sn 


2 

2 


1 6 
8 


19 

17 


21 


I 


7 

12? 2 


I 


0.085 Si 4- 5.02 Sn 


U 


I 


28 


32 


^ 


2O 


I 



notably those of the alloys of arsenic, 
antimony, bismuth, cobalt, and chro 
mium, the hydrogen treatment alone 



that were oxidized and then treated with 
hydrogen have been embrittled. The degree 
of the effect varies, and in general the 



a. o,2i per cent Zn; given treatment described under Fig. id. Type I embrittlement. Shows 
complete reduction of all zinc oxide at grain boundaries and within grains. 

b. 0.05 per cent B; same treatment as a. Type II embrittlement. Grain boundaries open as 
continuous cracks, reduction of oxide within grains uncertain. 

c. Traces Ce; same treatment as a. Type III embrittlement. Grain boundaries cracked open 
for short distances, elsewhere holes along grain boundaries. 

d. 0.05 per cent Ti; same treatment as a. Type IV embrittlement. Holes along grain boundaries 
rarely connected by cracks. 

e. o.io per cent Mg; same treatment as a. Type V embrittlement. Structure almost sound. 
/. 0.085 P er cent Si and 5.02 per cent Sn; same treatment as a. 



FREDERICK N. RHINES AND WILLIAM A. ANDERSON 



3*7 










JT IG> 2> VARIOUS ALLOYS OF COPPER. X 
[For descriptive legend see opposite page.] 



HYDROGEN EMBRITTLEMENT OF COPPER AND COPPER ALLOYS 



alloys containing oxides known to be easily 
reduced by hydrogen are most affected, 
while those with the most refractory oxides 
are embrittled to only a minor degree. 

These observations assume more definite 
significance when supported by the findings 
of the micrographic studies. Photomicro 
graphs of many of these alloys in the 
oxidized condition have been published. 4 
The similarity among the microstructures 
of the alternately oxidized and reduced 
samples makes the reproduction of a com 
plete pictorial record here unnecessary, but 
representative examples of the several gen 
eral types are presented in Fig. 2, a to e f , 
the various alloys are classified according 
to these type structures in Table i. The 
Roman numerals used for designating the 
structural types indicate the following: (I) 
grain boundaries open throughout oxidized 
zone, oxide within grains appears reduced; 
(II) grain boundaries open throughout oxi 
dized zone, reduction of oxide within grains 
uncertain because of small size; (III) holes 
and occasional channels at grain boundaries 
throughout oxidized zone; (IV) holes only 
at grain boundaries in oxidized zone; (V) 
no apparent discontinuities in the structure. 

Among the alloys in which the oxide 
precipitate was everywhere reduced and in 
which the embrittling effect was most ex 
treme are those containing indium, manga 
nese, nickel, phosphorus, tin, and zinc. 
There seems to be every reason to believe 
that the hydrogen has reacted vigorously 
with the foreign oxides in these alloys, pro 
ducing steam, which burst open the samples 
at the grain boundaries as in ordinary 
hydrogen embrittlement. 

The oxide in the tin alloy has been defi 
nitely identified by X-ray diffraction meth 
ods as SnC>2 and that in the phosphorus 
alloy tentatively by indirect means as 
P 2 05. Thus it is certain that in the tin 
alloys a complex oxide is not required for 
this behavior. Oxides in several other alloys, 
not included in the above group in which 
reduction was certain ; have also been 



identified; these include Si0 2) A1 2 3 , Cr 2 3 , 
Fe 3 O 4 , Pb 3 O 4 , Ti0 2? and ZrO 2 . Among the 
oxides not yet identified are several whose 
diffraction patterns are complex and do not 
correspond with those of any known oxide. 
These might be complex oxides in which 
copper occurs in addition to the alloying 
element. 

At the other extreme is a series of alloys 
that suffered no visible structural change 
during the hydrogen treatment subsequent 
to oxidation including columbium, gal 
lium, lithium, magnesium, and zirconium. 
Even among these, however, there has been 
a slight decrease in the resistance to the 
bend test, and it is altogether possible that 
there were holes at the grain boundaries too 
small to be distinguished from the oxide 
precipitate. 

Intermediate between these two groups 
is an almost continuous series in which 
embrittlement has definitely occurred, but 
to varying degrees. Least affected in this 
group are the alloys of aluminum, beryl 
lium, calcium, titanium, and perhaps iron; 
more affected are arsenic, bismuth, cerium, 
antimony, silicon, strontium, and tantalum; 
and still more affected are cobalt, chro 
mium, boron, and lead. 

The alloys of manganese, phosphorus, 
and silicon are of particular interest, for 
they show the influence of the amount of 
the alloy element. Since all were oxidized 
under identical conditions, those with the 
lowest alloy content will have oxidized to 
the greatest depth. The first alloy in each 
of these series that is, the one with the 
lowest amount of alloy addition has been 
oxidized throughout, and in the bend test 
the first crack and final rupture occurred 
together. With larger alloy additions, there 
remained in the sample a core of unoxidized 
tough metal that caused the sample to 
withstand several right-angle bends be 
tween the appearance of the first crack and 
the final break. The appearance of the 
series of manganese alloys after oxidation 
and reduction is shown in Fig. 3, 



FREDERICK N. RHINES AND WILLIAM A. ANDERSON 




FIG. 3. COPPER-MANGANESE ALLOYS. X 15* 
a, 0.033 per cent Mn; fe, 0.22 per cent Mn; c, 1.55 per cent Mn. , , . 

Annealed 45 hr. in copper plus copper oxide powder, 4 hr. in carbon monoxide and 2 hr in 
hydrogen without lowering of the temperature between treatments The cracks formed by hydro- 
gen embrittlement extend only to the limit of internal oxidation, the depth of which decreases as 
the alloy concentration increases. 



320 



HYDROGEN EMBRITTLEMENT OF COPPER AND COPPER ALLOYS 



The ternary alloys of zinc or tin with 
aluminum, beryllium or silicon all behaved 
similarly. Because of the high tin and zinc 
contents, these form a relatively thin sub- 
scale with a very heavy precipitate of the 
oxides. This zone was always thoroughly 
ruptured (Fig. 2/) by the hydrogen treat 
ment and often split away from the unoxi- 
dized core at the interface spontaneously 
before the bend test. For this reason, some 
of the samples presented a new and un 
affected surface and apparently were resist 
ant to the bend test. 

In a discussion of the internal oxidation 
studies carried out upon this series of alloys, 
A. A. Smith and J. S. Smart, Jr. 4 pointed 
out that the presence of an impurity such 
as iron would probably result in the pre 
cipitation of a second oxide in addition to 
the oxide of the intended addition. These 
alloys were analyzed for iron, and it was 
found that detectable quantities w r ere fre 
quently present, the quantities ranging 
from barely detectable traces to a maxi 
mum of 0.03 per cent in the alloys con 
taining beryllium and cerium. These 
experiments have shown iron to be defi 
nitely, though weakly, effective in causing 
susceptibility to hydrogen embrittlement. 
It is possible, therefore, that the slight 
embrittlement found among the beryllium 
and cerium alloys and tw r o or three of the 
others, including in particular vanadium, 
may result from the iron impurity. 

It is not unreasonable that a refractory 
oxide such as silica should be reduced by 
hydrogen in the presence of copper; indeed, 
it is well known that molten deoxidized 
copper will absorb silicon from a silica 
crucible. The copper simply provides a 
solvent for the liberated silicon, carrying 
it away from the reaction zone and thus 
permitting the reaction to continue with 
the accumulation of water until such a 
pressure is built up that an expanded cavity 
or open channel is blown in the copper. On 
this basis, it is to be anticipated that any 
oxide will be attacked to some extent, the 



less refractory ones reacting the most 
vigorously.* 

Pilling 5 has described the hydrogen em 
brittlement of oxygen-bearing copper- 
nickel alloys w r here a eutectic of copper and 
nickel oxides occurs in the cast structure, 
and Bengough and Hill 5 believed that a 
complex oxide of arsenic occurring in 
cast arsenical coppers leads to hydrogen 
embrittlement. 

The action of hydrogen on several of the 
unoxidized alloys deserves special notice. 
Varying quantities of oxides were inevitably 
introduced into the alloys when they were 
cast and their presence may well explain 
the unexpected embrittlement of the cal 
cium, cobalt and chromium alloys that 
exhibited typical cavities after the simple 
hydrogen anneal. The same explanation 
may apply also to the alloys of arsenic, 
antimony and bismuth annealed in hydro 
gen, but there is a further and perhaps more 
attractive explanation of the very pro 
nounced effects found in these. It is im 
probable that the similarity in behavior 
among the members of this familiarly asso 
ciated group of metals is accidental. Arse 
nic, antimony and bismuth each form 
several hydrides, among which are the 
well-known volatile compounds arsine, 
stibine and bismuthine. Such gases under 
pressure might well produce cracks and 
cavities in the copper, but there is as yet 
no proof of their formation in these alloys. 



* If the alloys are arranged in the order of the free 
energy of formation per gram atom of contained 
oxygen_ in their respective oxides, they appear in 
approximately the same sequence as when arranged 
according to their susceptibility to hydrogen em 
brittlement; those with the largest negative values of 
free energy correspond with the highest resistance to 
embrittlement. The free energies of some are as fol 
lows: CugO, -35150; MnO 2 , -51450; NiO, -51700; 
ZnO, -76000; SiO 2 , -95200; Cr 2 O 3 , -98300; TiOs, 
-106400; A1 2 O 3 , -119500; ZrO 2 , -122700; and 
MgO, 136370. The reaction is limited by the back 
pressure of the alloy element in solution in the copper, 
and _niuch less magnesium in solution would be 
required to stop the reduction of MgO than would be 
required for nickel and NiO. Thus refractory oxides 
surrounded by an alloy containing sufficiently large 
quantities of the corresponding metal, such as might 
appear in a casting of aluminum bronze with alumina 
inclusions, may be immune to reduction by hydrogen. 
Also, where large quantities of oxides are present 
their reduction may proceed to a certain limit and 
then cease. 



FRF.DKRK K N. KHINKS AND WILLIAM A, ANDERSON 



321 



Although the observations reported here 
often referred to individual tests, the work 
has all been examined for reproducibility, 
More than three hundred samples were 

tested, 

MKCHAMSM OK 



The relationship between the conclusions 
drawn from this research and the theory of 
the hydrogen embrittlement of copper as it 
is understood today may perhaps be more 
clearly demonstrated by briefly summariz 
ing the knowledge of the subject with the 
incorporation of the new observations* 

At temperatures higher than 4ooC,, hy 
drogen will diffuse into copper containing 
cuprous oxide and will reduce the oxide to 
metallic copper and water vapor, leaving 
the metal ruptured along the grain bound 
aries. 7 Hydrogen diffuses through copper 
much faster than the water vapor can 
escape by diffusion, and a pressure suffi 
cient to rupture the metal is thus built up, 8 
The cracks and holes apparently occur at 
the grain boundaries, because of inherent 
weakness there, because of more favorable 
conditions for nucleation there, because the 
oxide is more plentiful there, or because 
hydrogen enters more rapidly at the grain 
boundaries; on the other hand, it is con 
ceivable though less probable that rupture 
actually occurs randomly and the copper 
grains simply grow during annealing until 
limited by the discontinuities* Although 
several early investigations indicated that 
carbon monoxide and other reducing gases 
could embrittle copper, it is now known 
that hydrogen is the only gas (whether 
initially present in the gas or formed by 
reaction with water or hydrocarbons) that 
will produce this type of embrittlement in 
copper- 1 - 3 - 7 The oxygen may be present in 
any or all of several forms: cuprous oxide 
deposited as a eutectic mixture at the time 
of casting or precipitated from solid solu 
tion after an oxidizing anneal, oxygen in 
solid solution, or oxides of other metals 



embedded in the cop|>er. The effect is most 
pronounced in cast copper, where the oxide 
is most concentrated and easily reduced, 
and becomes less pronounced as the oxygen 
is more homogeneously distributed, as in a 
solid solution, or when it is combined in a 
less easily reduced form as in the oxide of 
some alloying element. In all cases the rate 
at first increases with the temperature, and, 
except in cast copper, which shows a 
maximum at 8ooC M * this increase con 
tinues at least to toooC. The log of the 
square of the depth of embrittlement in 
unit time is usually a linear function of the 
reciprocal of the absolute temperature. * 
The rate of embrittlement decreases gradu 
ally as the hydrogen pressure decreases; 
at very low pressures the rate of decrease 
is more rapid. 8 Oxygen-free or deoxidized 
copper is not susceptible to embrittlement 
upon annealing in hydrogen, 10 unless it has 
first been heated in an oxidizing atmos 
phere. 1 In this case, the rate of oxidation 
is much slower than that of the penetration 
of hydrogen, and except in very short re 
ducing treatments the depth of embrittle 
ment is controlled by the slower rate of 
internal oxidation. Hydrogen embrittle 
ment varies in its severity both in the depth 
of penetration of the effect and in the size 
and degree of continuity of the ruptures 
produced. Cuprous oxide or dissolved oxy 
gen may be withdrawn from copper to such 
an extent by heating in pure dry carbon 
monoxide 3 that the susceptibility to hydro 
gen embrittlement is destroyed; foreign 
oxides are but slowly removed, if at all, by 
this method. 

SUMMARY 

The new results obtained in this study 
are as follows: 

* This relationship has not been stated previously 
but can be derived from the data of Pilling,* Wymati,> 
and several others. The results of any single investi 
gator are usually self-consistent, but the discrepancies 
among the several rate studies reported in the litera 
ture suggest that the absolute value of the rate is 
subject to many variables, of which the distribution 
of the oxide is probably the most important. The 
decrease in the rate above 800 C. in cast copper has 
been ascribed to a welding of the cracks after the 
steam is released. 9 



322 



HYDROGEN EMBRITTLEMENT OF COPPER AND COPPER ALLOYS 



1. It has been demonstrated that very 
pure copper can be made susceptible to 
hydrogen embrittlement by saturating 
it with oxygen. Embrittlement follows 
whether the oxygen is held in solid solution 
or is allowed to precipitate as cuprous 
oxide. 

2. It has been demonstrated further that 
foreign oxides embedded in copper are 
reduced by hydrogen at elevated tempera 
tures and that a type of hydrogen em 
brittlement can result from this cause 
alone. 

3. The oxides known to be reduced in 
hydrogen, such as those of indium, manga 
nese, nickel, phosphorus, tin, and zinc, pro 
duce the most severe embrittlement, while 
the oxides of columbium, gallium, lithium, 
magnesium, and zirconium are only slightly 
effective; intermediate in their effect are 
the oxides of aluminum, arsenic, beryllium, 
bismuth, boron, calcium, cerium, cobalt, 
chromium, iron, lead, antimony, and 
titanium. 

4. A type of embrittlement less severe 
than hydrogen embrittlement is caused by 
the presence of a subscale on copper alloys. 
Subscales in which the oxide precipitate is 
heaviest or most concentrated at the grain 
boundaries are most damaging. 

5. Copper alloys containing small quan 
tities of arsenic, antimony, or bismuth were 
embrittled by annealing in hydrogen with 
out previous oxidation. It is suggested that 
volatile hydrides such as arsine, stibine, or 
bismuthine may have formed under suffi 
cient pressure to cause the rupture of the 
alloy. 

REFERENCES 

1. L. L. Wyman: A.I.M.E. Preprint (1931); Trans. 

A.I.M.E. (1933) 104, 141; (1934) in 305; 
(1940) 137, 291. 

2. S. B. Leiter: Trans. A.I.M.E. (1926) 73, 776. 

3. C. E. Ransley: Jnl. Inst. Metals (1939) 65, 147. 

4. F. X. Rhines: Trans. A.I.M.E. (1940) 137, 246. 

5. N. B. Pilling ://. Franklin Inst. (1918) 186,373. 

6. G. D. Benguugh and B. P. Hill; Jnl. Inst. Metals 

(ipio^ 3, 34- 

7. E. Heyn: Ztsch, Ver. dcut. Ing (1900) 44, 503, 

and others. 

8. N*. B. Pilling: Trans. A.I.M.E. (1919) 60, 322. 

9. C, S. Smith and C. R. Hay ward : Jnl. Inst. Metals 

(19261 36> 2H. 

10. H. Moore and S. Beckinsale: Jnl. Inst. Metals 
(1921) 25* 2i9i and others. 



DISCUSSION 

(ir. //. Bassett, Jr., presiding) 

A. PHILLIPS* AND D. B. GRAVES, f New 
Haven, Conn. The production and increasing 
use of substantially oxygen-free copper has 
converted one academic query into a question 
of practical significance; namely, whether 
oxygen present in solid copper within the 
solubility limits renders the copper susceptible 
to hydrogen embrittlement. A study of this 
matter has been made at the Hammond 
Metallurgical Laboratory and the results to 
date substantiate the claim of Rhines and 
Anderson that embrittlement may be encoun 
tered without the presence of discrete particles 
of cuprous oxide. 

In order to obtain conclusive evidence 
regarding the susceptibility to gassing of copper 
holding oxygen in solid solution, it is of course 
imperative to preclude the possibility of 
precipitation of Cu^O after the saturation 
treatment and prior to, or during, the hydrogen 
treatment. This necessity has been properly 
appreciated by Rhines and Anderson and there 
is no reason to believe that precipitation of 
oxide had taken place within the specimens 
during the sequence of treatments. The authors 
did, however, hydrogen-anneal their specimens 
with a surface layer of oxide produced by 
heating in the copper-cuprous oxide mixture. 
As will be noted later, the present writers 
removed the surface oxide produced by the 
saturation treatment in order to eliminate any 
possible concomitant gassing effects at the 
surface. 

The sequence of treatments observed at the 
Hammond Laboratory may be summarized as 
follows : 

1. Oxidized surface in air to produce a 
visible coating. 

2. Attained saturation by heating for 42 hr. 
in purified nitrogen. 

3. Specimens quenched to prevent precipita 
tion of cuprous oxide. 

4. Scale removed by solution in dilute nitric 
acid. 

5. Reheated in pure nitrogen to a tempera 
ture above the saturation temperature. 



* Professor of Metallurgy, Yale University, 
t Graduate Student, Department of Metallurgy, 
Yale University. 



DISCUSSION 



3 2 3 



6. Temperature adjusted to gassing temper 
ature and specimen held at the temperature to 
ensure equilibrium conditions. 

7. Passed in hydrogen to replace nitrogen. 
Heated in hydrogen atmosphere for 2 hours. 

8. Hydrogen replaced by nitrogen and speci 
men allowed to cool. 

The treatments were made on tensile-test 
specimens in order to establish a quantitative 
basis for studying hydrogen-embrittling effects. 
A typical set of our experimental results has 
been selected (Table 2) to confirm Rhines and 
Anderson s findings. It -is to be noted that in 
the cases cited the hydrogen treatments were 
also carried on at temperatures above the 
saturation temperature and for this reason 
the results can in no way be attributed to 
the precipitation of cuprous oxide. 

TABLE 2. Typical Results of Experiments 
by Phillips and Graves 



Saturation 
Temperature, 
Deg. C. 


Temperature of 
Hydrogen Treat 
ment (2 Hr.) r 
Deg. C. 


Tensile 
Strength, 
Lb. per 
Sq. In. 


Elonga 
tion in 2 
In., Per 
Cent 


900 
ooo 


No hydrogen 
treatment 
935 


31,000 
31,000 
14,500 


45 
39 
9. o 


900 
900 


945 
850 


14,800 
14,300 
15,000 
17,000 


8-5 
8.5 
9-5 
8-5 


900 


Slowly heated to 


15,000 
8,100 


7-5 
5- 5 


No saturation 
treatment 
(OFHC) 


900 in hydrogen 
and cooled in 
hydrogen 
Slowly heated to 
900 in hydrogen 
and cooled in 
hydrogen 


7,700 

29,900 
29,200 


5-5 

34- S 
34-5 



It is interesting to note the marked embrittle- 
ment of the specimens simply heated to gooC. 
in hydrogen (after saturation) and allowed to 
cool in hydrogen. No surprise is occasioned, of 
course, by the observation that the treatment 
encouraging precipitation of oxide was asso 
ciated with the maximum embrittlement, but 
rather by the evidence that gassing of copper 
containing only about 0.003 P er cent oxygen 
can lower the strength from some 31,000 Ib. 
per sq. in. to 8000 Ib. and the elongation from 
over 40 per cent to 5 per cent. 

S. ROLLE,* New York, N. Y. Some years 
ago the Bell Telephone Laboratories did con 



siderable work on the hydrogen embrittlement 
of copper. As a result of their studies they 
standardized on &5oC. as the optimum 
temperature for heating copper in hydrogen 
to produce most rapid and pronounced embrit 
tlement. I wonder whether temperatures above 
and below this were carefully explored and, if 
so, whether at elevated temperatures any 
healing action on fractures was noted. 

E. E. SCHUMACHER,* New York, N. Y. 
In the latter part of the ig2o s we were inter 
ested in the use of an oxygen-free copper as a 
conductor in submarine cable. At that time a 
comprehensive study was made of the action 
of hydrogen on hot copper in connection with 
an investigation of means of producing oxygen- 
free or deoxidized copper. The degree of 
embrittlement by hydrogen of the various 
coppers tested was evaluated by the number of 
bends required to produce failure, together 
with tensile and ductility characteristics, after 
the specimens had been heated in hydrogen at 
800 to 85oC. for periods of K to i hr., depend 
ing upon the size of the specimen. The tem 
perature of 800 to 85oC. was chosen because 
of the following considerations: Smith and 
Hayward, in 1926, and others had reported 
that the embrittling action of hydrogen on 
copper containing oxygen became pronounced 
at temperatures of 500 to 6ooC., had its 
maximum effect in the region from 700 to 
about 8soC., and for temperatures above 
approximately QOO C C. structural repair was 
initiated, which progressively increased in 
rate to the melting point of the copper. The 
rate of growth of the reduced zone for low- 
oxygen coppers, which were of interest to us, 
increased as the temperature was increased. 
These reported findings were spot-checked by 
us and confirmed. Consequently we chose a 
temperature for the embrittling treatment at 
which the penetration would be most rapid, but 
below the temperature at which the healing 
action commenced. The result was a choice of 
a temperature range of 800 to 85oC. 

In answer to Mr. Rolle s question, we did 
observe a healing action at temperatures above 
85oC. Specific results obtained on 0.03 2-in. 
dia. copper wire containing 0.03 per cent 
oxygen were as follows: Number of right-angle 



* Assistant Manager, Scomet Engineering Co. 



* Research Metallurgist, Bell Telephone Labora 
tories. 



HYDROGEN EMBRITTLEMENT OF COPPER AND COPPER ALLOYS 



bends necessary to produce failure: (i) annealed 
in air ioooC. for i hr., 9 bends, (2) annealed in 
hydrogen 8ooC. for i hr., rj^ bend, (3) 
annealed in hydrogen 9ooC. for i hr., 3 bends, 
(4) annealed in hydrogen ioooC. for i hr., 
6 bends. The treatment at the higher tempera 
tures, where the embrittlement was less 
marked than at 85oC. 3 resulted in considerable 
recrystallization and grain growth in the copper 
with the formation of new tight boundaries. 

It was observed also that copper that had 
been treated with hydrogen at ioooC. had 
sufficient ductility to permit subsequent cold- 
drawing and annealing. Following such a 
practice, a deoxidized copper with excellent 
ductility and freedom from embrittlement by 
reducing gases was produced. 

S. ROLLE. Our laboratory confirmed Mr. 
Schumacher s findings but not entirely with 
regard to the ductility of copper annealed in 
hydrogen at ioooC. We found that this treat 
ment "fully healed 11 the metal but its ductility 
was lower than that of metal that had not been 
so abused, As the result of Mr. Schumacher s 
and our investigation, we have standardized on